DJI Mini 5 Pro vs Mini 4 Pro - comparison of photo quality

You are currently viewing DJI Mini 5 Pro vs Mini 4 Pro - comparison of photo quality

As I men­tioned in my arti­cle “DJI Mini 5 Pro – Pho­to Qual­i­ty”, I have now pur­chased the new DJI Mini 5 Pro. The main rea­son for the upgrade was the new 1-inch cam­era sen­sor, which I hope will fur­ther improve the qual­i­ty of the pho­tos. But is that real­ly the case?

As in my pre­vi­ous drone tests, I once again took com­par­i­son pho­tos with iden­ti­cal set­tings using both drones from my win­dow. Unfor­tu­nate­ly, the sky was over­cast this time.

Both drones were man­u­al­ly set to ISO 100 and an expo­sure time of 1/200s. Focus was set via the touch­screen on the gable of the house with the orange clink­er bricks. The pho­tos were tak­en simul­ta­ne­ous­ly at the high­est res­o­lu­tion (50 and 48 megapix­els, respec­tive­ly) to achieve opti­mal com­pa­ra­bil­i­ty. I ignored the min­i­mal dif­fer­ence in lens aper­ture (Mini 5 Pro f/1.8 vs. Mini 4 Pro f/1.7), which cor­re­sponds to only about 1/10 of a f-stop. The DNG files were then import­ed into Light­room Clas­sic V 14.5.1 and edit­ed min­i­mal­ly with the fol­low­ing iden­ti­cal settings:

I left all oth­er options in Light­room at their default settings.

Overview

Here are both images in direct comparison:

On the left (or at the top when viewed on a mobile phone), you can always see the image from the Mini 5 Pro, and on the right that of the Mini 4 Pro. The direct com­par­i­son already clear­ly reveals that the Mini 5 Pro has a vis­i­bly larg­er field of view.

Light­room Clas­sic uses a lens pro­file inte­grat­ed into the DJI Mini 4 Pro DNG file to cor­rect dis­tor­tion and chro­mat­ic aber­ra­tions in images cap­tured by the Mini 4 Pro. This can­not be dis­abled in Lightroom:

The DNG file of the Mini 5 Pro, on the oth­er hand, does not con­tain an inte­grat­ed lens pro­file and appar­ent­ly does not need one. Either the optics of the Mini 5 Pro have already been cor­rect­ed to such an extent that dig­i­tal dis­tor­tion cor­rec­tion is no longer nec­es­sary, or (much more like­ly!) the cor­rec­tion of the DNG file is instead already per­formed in the Mini 5 Pro by the more pow­er­ful CPU that is like­ly inte­grat­ed there.

The size of the DNG files is also inter­est­ing: the file­size of the Mini 5 Pro is 71 MB, while the Mini 4 Pro file is as large as 96 MB.

How­ev­er, the size of a DNG file only pro­vides lim­it­ed infor­ma­tion about image qual­i­ty. Much depends on the effi­cien­cy of the com­pres­sion algo­rithms. For exam­ple, the CR3 RAW files from my Canon EOS R5 Mark II are only between approx­i­mate­ly 45 and 60 MB in size, depend­ing on the ISO set­ting and sub­ject. As a gen­er­al rule, files become larg­er with increas­ing noise at high­er ISO val­ues and more details in the scene. Noise in par­tic­u­lar can­not be com­pressed effectively.

Image adjustment

Using iden­ti­cal devel­op­ment set­tings, the Mini 5 Pro’s image appears slight­ly dark­er and warmer than that of the Mini 4 Pro. This is like­ly par­tial­ly due to the slight­ly wider aper­ture of the Mini 4 Pro. To make the fol­low­ing exam­ples eas­i­er to com­pare, I there­fore adjust­ed the devel­op­ment set­tings for the Mini 4 Pro image slight­ly to achieve the best pos­si­ble match:

Com­pared to the Mini 5 Pro set­tings, I increased the col­or tem­per­a­ture by 300 °C and the tint by 10, and reduced the expo­sure by 0.4 EV. Here is the result in a direct comparison:

So both images now look quite com­pa­ra­ble in terms of bright­ness and col­or, at least in the overview.

Comparison of details

So it’s time to go into the details. How does the image sharp­ness com­pare in direct comparison?

To do this, I com­pared sev­er­al sec­tions of the images above. To make the dif­fer­ences clear­ly vis­i­ble even on small mobile dis­plays, I enlarged the sec­tions to 400% in Light­room’s com­pare view. As before, the image from the Mini 5 Pro is on the left and the image from the Mini 4 Pro is on the right:

As can be seen in the three com­par­i­son images, the images on the left tak­en with the Mini 5 Pro are actu­al­ly all slight­ly sharp­er, high­er in con­trast, and low­er in noise than those tak­en with the Mini 5 Pro. How­ev­er, the dif­fer­ence is not very large and is only clear­ly vis­i­ble at this high magnification.

Dynamic range

Sharp­ness is only one cri­te­ri­on for the qual­i­ty of a lens/sensor com­bi­na­tion. The dynam­ic range of a sen­sor, i.e., its abil­i­ty to dif­fer­en­ti­ate between very bright and very dark parts of a scene in an image, is also very impor­tant. The dark­er the areas in the sub­ject become, the low­er the sig­nal-to-noise ratio of the sen­sor pix­els in ques­tion and the stronger the image noise becomes. Here, I expect­ed a vis­i­ble improve­ment with the Mini 5 Pro due to its larg­er sen­sor and larg­er indi­vid­ual pixels.

Basics

In very bright areas of an image, the sen­sor pix­els will at some point become sat­u­rat­ed, or “ful­ly charged,” so to speak. Any addi­tion­al pho­tons will then have no effect. It is like a full bar­rel that over­flows at some point. Cor­rect­ing over­ex­po­sure in dig­i­tal sen­sors in post-pro­cess­ing is there­fore only pos­si­ble to a very lim­it­ed extent.

The sit­u­a­tion is dif­fer­ent in dark areas. Here, depend­ing on the qual­i­ty of the sen­sor, exten­sive cor­rec­tions are still pos­si­ble. For this rea­son, par­tic­u­lar atten­tion should be paid to the bright, rel­e­vant parts of the sub­ject dur­ing expo­sure (ETTR - Expose to the right) in order to make opti­mal use of the sen­sor’s dynam­ic range.

How­ev­er, the dark­er the areas in the sub­ject become, the low­er the sig­nal-to-noise ratio of the rel­e­vant sen­sor pix­els and the greater the image noise. Here, I expect­ed a vis­i­ble improve­ment with the Mini 5 Pro due to the larg­er sen­sor and larg­er indi­vid­ual pixels.

To visu­al­ize the dif­fer­ences, I took test shots with both drones that were under­ex­posed by 2 2/3 f-stops. To do this, I reduced the expo­sure time to 1/1250 while keep­ing all oth­er set­tings the same. The DNG files cre­at­ed in this way ini­tial­ly appear very dark in Light­room when using the same devel­op­ment set­tings as for the cor­rect­ly exposed images shown above:

When they are bright­ened by 2.66 f-stops each using the expo­sure slid­er in Light­room, they ini­tial­ly look in the overview as they did in the cor­rect­ly exposed shots above:

Below, I am pre­sent­ing the same crops shown above, again at 400% mag­ni­fi­ca­tion, for direct com­par­i­son. As before, the left crop shows the crops from the Mini 5 Pro, while the right crops show those from the Mini 4 Pro:

As expect­ed, both images now show sig­nif­i­cant noise. In a direct com­par­i­son, how­ev­er, I think the Mini 5 Pro per­forms bet­ter again, cap­tur­ing more detail. It is notice­able, how­ev­er, that the Mini 5 Pro has sig­nif­i­cant­ly more col­or noise than the Mini 4 Pro. This is par­tic­u­lar­ly evi­dent in the church win­dow, for example.

How­ev­er, the rea­son for this seems to be that the Mini 4 Pro aggres­sive­ly removes col­or noise inter­nal­ly, result­ing in all images show­ing sig­nif­i­cant­ly less col­or detail. This can be seen very clear­ly in the last image tak­en with the Mini 4 Pro, in which all the tiles have almost the same col­or. The Mini 5 Pro con­tin­ues to show the dif­fer­ent red/orange/brown tones that are present.

In the above com­par­isons, it is also worth not­ing that due to the high­er lens aper­ture of the Mini 4 Pro , the expo­sure of the under­ex­posed Mini 4 Pro file in Light­room only had to be increased by 2.26 EV, while that of the Mini 5 Pro had to be increased by as much as 2.66 EV. Despite the stronger adjust­ment, the Mini 5 Pro exhibits bet­ter noise performance.

Although col­or noise is indeed more pro­nounced with the Mini 5 Pro, the images cap­ture the actu­al col­ors of the sub­ject much bet­ter. With the Mini 4 Pro, the col­ors are severe­ly bleached. In this com­par­i­son, I much pre­fer the col­or rep­re­sen­ta­tion of the Mini 5 Pro. Fur­ther­more, col­or noise can be removed rel­a­tive­ly well in post-processing.

AI denoising

But we live in the age of AI. For some time now, Adobe Light­room offers an inte­grat­ed noise reduc­tion fea­ture with AI algo­rithms. I was curi­ous to see what could be done with the noisy images above. So I reworked the under­ex­posed drone images with the default set­ting of 50 using the Detail | Denoise menu item in the Develop menu:

Here you can see the denoised crops:

Once again, it’s amaz­ing what AI can still get out of noisy images. In a direct com­par­i­son, how­ev­er, I still like the results from the Mini 5 Pro a lit­tle bet­ter. In par­tic­u­lar, the mason­ry around the church win­dow and also in the last exam­ple shows more details with the Mini 5 Pro, while the struc­tures are more blurred with the Mini 4 Pro, some­times with notice­able artifacts.

Note on my workflow

cI would like to express­ly point out that what is writ­ten here refers to my own work­flow with Adobe Light­room Classic.

Oth­er RAW con­vert­ers may pro­duce dif­fer­ent results. How­ev­er, I think it is unlike­ly that there will be fun­da­men­tal dif­fer­ences from my results. For those who would like to try it out for them­selves, I have made the DNG files used in this com­par­i­son avail­able for down­load in a ZIP file so that you can do your own tests:

DNG-Files Mini 5 Pro vs Mini 4 Pro (Cau­tion, 258MB download!)

Addi­tion­al DNG test files - includ­ing those from the Mini 3 Pro, Air 3s, iPhone 15 Pro, and my Canon EOS R5 Mark II - can be found at the end of my arti­cle on the Air 3s. Unfor­tu­nate­ly, despite the same motif, direct com­par­i­son with the Mini 4/5 Pro images shown here is dif­fi­cult because the weath­er was sig­nif­i­cant­ly bet­ter and the light was much harsh­er at the time.

Resumée

As good as the images tak­en with the Mini 4 Pro are when using the cor­rect tech­nique and post-pro­cess­ing, the images tak­en with the Mini 5 Pro are even bet­ter. In par­tic­u­lar, the dynam­ic range has improved thanks to the larg­er sen­sor. How­ev­er, the dif­fer­ences are not huge and are actu­al­ly only vis­i­ble at high mag­ni­fi­ca­tions. The AI noise reduc­tion also enables a sig­nif­i­cant improve­ment in the devel­op­ment result for both cameras.

But the Mini 5 Pro also offers sev­er­al addi­tion­al fea­tures over the Mini 4 Pro, which I have already described in detail in my pre­vi­ous report, “DJI Mini 5 Pro – Pho­to Qual­i­ty”. I would par­tic­u­lar­ly like to men­tion the free panora­ma mode, which I have already used fre­quent­ly and enjoyed using on its big sis­ter, the DJI Air 3s.

Nev­er­the­less, the Mini 4 Pro remains a won­der­full drone that I have used fre­quent­ly and enjoyed using so far. But as is often the case, the bet­ter is the ene­my of the good - so I upgrad­ed to the Mini 5 Pro.

What do you think?

This Post Has 19 Comments

  1. andrew

    you are con­fused with GB and MB 🙂

    1. Admin

      Ooops, you’re right, thanks for point­ing that out, I’ve cor­rect­ed it

  2. Jim Lamont

    Thank you. I agree that someAI-tech­niques for enlarg­ing raise sig­nif­i­cant ques­tions about the real­i­ty of the result­ing pho­to­graph. How­ev­er, I was using « Bicu­bic Smoother » which does not seem to pose such a prob­lem. Again, when I uprezzed the 12 mpx image using Bicu­bic I could see lit­tle or no dif­fer­ence from the 50 mpx. in fact the uprezzed 12 mph image some­times looked crisper with slight­ly more detail. And it is not as if the 50 mpx cap­ture comes with­out cost: 50mpx cap­tures with the mini 5 have very slight­ly low­er dynam­ic range and poor­er colour accord­ng to sev­er­al sources, nei­ther of which can be recov­ered in post-processing.I am, like you, try­ing to estab­lish a best tech­nique so this issue is very impor­tant to me. I very much respect your work, so I ask you: do you have exper­i­men­tal proof that 50 is actu­al­ly bet­ter than 12 for the mini 5.? I am sug­gest­ing that what may seem obvi­ous may not be sup­port­ed empir­i­cal­ly in this case. What do you think?

    1. Admin

      I have nev­er test­ed upscal­ing myself in com­par­i­son to native res­o­lu­tion, as it seems phys­i­cal­ly illog­i­cal to me. The 12-megapix­el res­o­lu­tion is cap­tured with the same pix­els as the 50-megapix­el res­o­lu­tion, the only dif­fer­ence being that 4 pix­els are already inter­con­nect­ed inter­nal­ly on the chip and processed as a sin­gle pix­el. Since more area is avail­able, this improves both noise per­for­mance and dynam­ic range. How­ev­er, the same effect can also be achieved in post-pro­cess­ing. Inci­den­tal­ly, dpre­view once made a very inter­est­ing video on this top­ic.
      Best regards,
      Gerd-Uwe

      1. Jim Lamont

        I agree it seems phys­i­cal­ly illog­i­cal at first glance . How­ev­er the 50 mpx are not in a bay­er sen­sor arrange­ment when looked at as a whole. So inter­pret­ing them as 50 sep­a­rate pix­els comes at a cost, notably poor­er colour con­tent nfor­ma­tion. The results you cite for bay­er sen­sors are not entire­ly valid for quad bay­er sen­sors. Any­way, if you try the exper­i­ments you will see the evi­dence with your own eyes.

  3. perwea

    Good work, tank you!
    A low light com­par­i­son would have been won­der­ful. IMHO bet­ter than sim­u­lat­ing avail light by doing adjustments.

  4. Jim Lamont

    I have done some test­ing of 12 vs 50 mps cap­tures. I can pro­vide details, but I found no con­sis­tent improve­ment using 50. I have been try­ing to get more pix­els in a giv­en area (rather than the usu­al, more pix­els with more area). I want to get 300 ppi for 20x30 inch prints, of small­ish areas shot ver­ti­cal­ly often.

    1. Admin

      Hel­lo again, Jerome,
      50 megapix­els should be suf­fi­cient for a 20x30 inch print with­out any prob­lems. I have already print­ed this size sev­er­al times. How­ev­er, due to the noise sus­cep­ti­bil­i­ty and low­er dynam­ic range of the small sen­sor com­pared to a sys­tem cam­era, it makes sense to post-process the DNG files inten­sive­ly. I almost exclu­sive­ly use 5x expo­sure series on the DJI drone, which I then merge into an HDR in Adobe Light­room. This already sig­nif­i­cant­ly improves the noise and dynam­ic range. The rest can then be removed with Light­room’s AI denois­ing if necessary.
      Best regards,
      Gerd-Uwe

      1. Jim Lamont

        My prob­lem is more dif­fi­cult than mere­ly get­ting 50 Mpx. Get­ting more Mpx by sim­ply increas­ing the area of the com­po­si­tion is easy. I want to com­pose a shot that is 12 Mpx, and then, for that same com­po­si­tion, get 50 Mpx (good pix­els, not the faux quad bay­er 50 Mpx). Long ago I did some­thing sim­i­lar with my D2Xs: in order to get more pix­els for print­ing large: when I had a com­po­si­tion I real­ly liked I shot it a sec­ond time with a 3-frame panora­ma using a larg­er focal length.. With the essen­tial­ly sin­gle focal length of the mini 5 this is more difficult.

        1. Admin

          You can also increase the res­o­lu­tion of an image dur­ing post-pro­cess­ing by sim­ply tak­ing sev­er­al shots in quick suc­ces­sion and then load­ing them as lay­ers in Pho­to­shop, as an exam­ple. Due to the inevitable min­i­mal move­ment of the drone between shots, these indi­vid­ual images have a slight off­set, so that indi­vid­ual details of the sub­ject are mapped to dif­fer­ent pix­els. If you then align the lay­ers in Pho­to­shop, com­bine them into a smart object, and ren­der them using the “medi­an” option, the res­o­lu­tion is increased and noise is reduced. Cur­rent smart­phones work in a sim­i­lar way inter­nal­ly. I see, I’ll have to write an arti­cle about that sometime.

          1. Jim Lamont

            Thank you. If that « hyper res­o­lu­tion » approach works that would be won­der­ful. Unfor­tu­nate­ly it does not seem to work. I tried it sev­er­al times today with 20 frames shot n quick suc­ces­sion and tried two approach­es to stack­ing and com­bin­ing them (as the medi­an and as lay­ers with decreas­ing opac­i­ty after uprez­zng) , in strict accor­dance with sev­er­al web sources for hyper res­o­lu­tion tech­niques. Nei­ther approach increased the per­cep­ti­ble detail to any sig­nif­i­cant degree, if at all. Both decreased the noise. The medi­an approach actu­al­ly made the image very slight­ly blur­ri­er, which makes sense when you think about it. Both approach­es per­mit­ted rather more sharp­en­ing which made com­par­isons dif­fi­cult. How­ev­er it was very obvi­ous that nei­ther approach increased the actu­al detail to any sig­nif­i­cant degree, if at all. Cer­tain­ly noth­ing even close to what I can achieve by comb­ing three frames shot with a larg­er focal length with my con­ven­tion­al cam­era. Have you achieved dif­fer­ent results. If so, pre­cise­ly how? Thank you again.

          2. Admin

            Hel­lo Jim,
            com­bin­ing mul­ti­ple indi­vid­ual images into a sin­gle image has long been com­mon prac­tice in astropho­tog­ra­phy (stack­ing). In fact, it serves in par­tic­u­lar to reduce noise. In addi­tion, with slight­ly off­set images, indi­vid­ual pix­els end up on dif­fer­ent­ly colour-sen­si­tive pix­els, so that colour noise, which is often very dis­turb­ing with Quad-Bay­er sen­sors, is also reduced. The com­bined images nat­u­ral­ly have the same res­o­lu­tion, but appear sharp­er due to the reduced grey and colour noise. How­ev­er, there are also prob­lems with stack­ing. Cor­rect­ed wide-angle lens­es have a rel­a­tive­ly longer focal length and mag­ni­fi­ca­tion at the edges to com­pen­sate for the oth­er­wise unavoid­able bar­rel dis­tor­tion. Only fish­eye lens­es do not have this. If you then shift the image by one pix­el in the cen­tre, for exam­ple, the edge shifts by two pix­els. Con­gru­ent stack­ing is then no longer pos­si­ble and the images become blurred towards the edges. In addi­tion, the opti­cal qual­i­ty of the Mini 5 Pro’s small lens is nat­u­ral­ly lim­it­ed. What­ev­er you do, you can try to get the best out of it with post-pro­cess­ing tech­niques, but you will nev­er achieve the qual­i­ty of a sys­tem cam­era with a good fixed focal length (which costs many times more than the Mini 5 Pro). I com­pared my EOS R5 and my iPhone 15 Pro in my arti­cle ‘DJI Mini 4 Pro - pho­to qual­i­ty in com­par­i­son’.
            Best regards,
            Gerd-Uwe

          3. Jim Lamont

            Thank you for your infor­ma­tive reply. It con­firms my expe­ri­ence that hyper res­o­lu­tion tech­niques do not work with the mini 5 to increase res­o­lu­tion, although they can be used to reduce noise. I am look­ing for­ward to read­ing your post com­par­ing iPhone to reg­u­lar cam­era. I should say here that I real­ly appre­ci­ate your analy­ses. You are the only per­son I know doing seri­ous inves­ti­ga­tions of drone photography.

  5. jerome

    Hi,
    Thank you for this comparison.
    Maybe you should have com­pared pho­tos tak­en at 12mpx too as it’s the native res­o­lu­tion of both sensors ?

    1. Admin

      Hi Jerome,
      the native res­o­lu­tion of the sen­sor of the Mini 5 Pro is 50mpx. But unlike con­ven­tion­al sen­sors for still cam­eras, this sen­sor uses the Quad Bay­er for­mat. In 12mpix mode, 4 pix­els of the same col­or are then inter­con­nect­ed with­in the chip, so that it then seems to be a 12mpix chip. 

      I always use the high­est avail­able res­o­lu­tion; even my old Canon EOS 5D already had 12 megapix­els 20 years ago. That res­o­lu­tion has not been enough for me for a long time. I pub­lish my pic­tures in cal­en­dars in for­mats up to DIN A2 and pro­duce large prints. But I think that you get the same result as in 12mpx mode any­way if you sim­ply down­scale the 50mpx image to 1/4 dur­ing post-pro­cess­ing. In this case, too, the four indi­vid­ual high-res pix­els are added togeth­er, which also results in less noise. Down­scal­ing reduces noise very effectively.

    2. Jim Lamont

      Thank you for your very inter­est­ing arti­cle. I too would appre­ci­ate a com­par­i­son between the 12 and the 50 max modes. I have done some test­ing myself, shoot­ing ver­ti­cal low alti­tude shots, and the dif­fer­ence seems to vary. Some­times 50 max is very slight­ly bet­ter when viewed at 100% and sharp­ened (it can ben­e­fit from more sharp­en­ing and increased clar­i­ty). How­ev­er some­times the 12 max appears bet­ter. All shots in excel­lent light, at 100 iso and speeds of 1/1000 or faster. Puz­zling. I would also be very inter­est­ed in see­ing a study of how to increase the total pix­els in a giv­en area using free panora­ma., that is not just get­ting more pix­els by shoot­ing a big­ger area. my web­site is http://www.jalamont.ca. I share your inter­est in high qual­i­ty drone photography.

      1. Admin

        Hel­lo Jerome,
        I don’t think it makes sense to test the 12-megapix­el mode sep­a­rate­ly. Since the 12 megapix­els on the sen­sor are sim­ply cal­cu­lat­ed by com­bin­ing 4 pix­els of the 50-megapix­el sen­sor, you can achieve the same result by sim­ply reduc­ing the res­o­lu­tion to 25% in post-pro­cess­ing. Of course, the 12-megapix­el images look bet­ter in 100% view than the 50-megapix­el images in 100%. How­ev­er, it is impor­tant to keep in mind that the 50-megapix­el image is four times larg­er. For com­par­i­son, you would have to view the 50-megapix­el image at 25%, and then you will no longer see any dif­fer­ence between the two.
        Best regards,
        Gerd-Uwe

        1. Jim Lamont

          Yes, I agree that con­vert­ing a 50 Mpx cap­ture to 12 will give bet­ter noise char­ac­ter­is­tics and so on. Sim­i­lar­ly, and per­haps sur­pris­ing­ly, uprezzing a mini 5’s 12 Mpx to 50 Mpx will give a result that is very often indis­tin­guish­able from a 50 Mpx quad bay­er cap­ture. In fact the 50 Mpx cap­ture can be worse, with poor­er colour, low­er local con­trast, and low­er dynam­ic range. What I do not under­stand is the con­di­tions that make cap­tures at 50 Mpx worse than 12 and when better.

          1. Admin

            If you use AI to enlarge a 12-megapix­el image to 50 megapix­els, it may look sharp­er than a ative 50-megapix­el image, but the vis­i­ble sharp­ness has noth­ing to do with the orig­i­nal sub­ject. Instead, the AI invents addi­tion­al pix­els that fit well. I rec­om­mend read­ing my arti­cle AI denois­ing - fact or fake? on this top­ic. Although it deals with AI denois­ing, what is described there also applies to enlarge­ment. Feel free to try it out for your­self: Take a pic­ture of a news­pa­per page from a dis­tance with 12 and 50 megapix­els, enlarge the 12-megapix­el image to 50 megapix­els, and com­pare the two 50-megapix­el images with each oth­er. The enlarged image may appear sharp­er, but the one tak­en with 50 megapix­els will be eas­i­er to read. You have to decide for your­self whether you want fact or fake 😉

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.