Blog

As already men­tioned in my first short report on the Mini 4 Pro, I have now pur­chased the new Mini 4 Pro, which will replace my pre­vi­ous Mini 3 Pro. After a first test flight, I was par­tic­u­lar­ly inter­est­ed in the pho­to qual­i­ty, since I use my drones main­ly for pho­tog­ra­phy. As a pas­sion­ate pho­tog­ra­ph­er, I exclu­sive­ly use the RAW / DNG for­mat in the high­est sen­sor res­o­lu­tion of 48 megapixels.

Accord­ing to the tech­ni­cal spec­i­fi­ca­tions, the cam­era of the Mini 4 Pro has not changed com­pared to its pre­de­ces­sor Mini 3 Pro. Although it looks the same on the out­side, it is not: my ND fil­ters from the Mini 3 Pro do not fit to it.

How­ev­er, DJI points out that the image qual­i­ty of the Mini 4 Pro is still improved due to the more pow­er­ful dig­i­tal image pro­cess­ing. I already had a sim­i­lar impres­sion with my larg­er drone, the DJI Air 3, which also uses a 1/ 1.3-inch sen­sor and optics with the same spec­i­fi­ca­tions in its wide-angle cam­era. The com­par­i­son between Air 3 and Mini 3 Pro can be found here.

While many peo­ple think that RAW files con­tain only the actu­al, unprocessed sen­sor data, this is not the case. In fact, the data is always pre-processed in your cam­era - more so with some cam­eras, less so with oth­ers. Arti­facts that occur dur­ing this process are very dif­fi­cult to remove lat­er, even with the best RAW devel­op­ment software.

A par­tic­u­lar prob­lem aris­es from the quad-bay­er sen­sor cho­sen by DJI. Com­pared to sen­sors with a clas­sic Bay­er design, the col­or res­o­lu­tion of these is reduced to a quar­ter and thus only cor­re­sponds to that of a 12-megapix­el sensor.

I have already writ­ten more about the Bay­er / Quad Bay­er top­ic in my arti­cle about the Mini 3 Pro, so I don’t want to repeat every­thing here.

How­ev­er, since all RAW pro­cess­ing tools I know are designed for the stan­dard Bay­er design, the DJI drones already con­vert the raw data into a so-called lin­ear DNG for­mat. In this process, the col­or map­ping of the indi­vid­ual pix­els (the “de-Bay­er” process) is already done inter­nal­ly in the drone. For this pur­pose, the assumed col­or of each indi­vid­ual sen­sor pix­el is derived from the sur­round­ing RGB pix­els accord­ing to a pro­pri­etary algo­rithm. Anoth­er dis­ad­van­tage of lin­ear DNG files is that they con­tain the data of all three col­or chan­nels for each sen­sor pix­el, which results in three times larg­er file size than, for exam­ple, the RAW file of the Canon EOS R5 pri­or to the de-Bay­er process.

A clos­er look at the DNG files reveals that post-pro­cess­ing (denois­ing / sharp­en­ing) has obvi­ous­ly tak­en place here as well. The Mini 3 Pro in par­tic­u­lar shows quite notice­able sharp­en­ing arti­facts. These are not as pro­nounced in the Air 3, which was released later.

Computational photography

The image qual­i­ty of dig­i­tal cam­eras there­fore depends not only on the optics and the sen­sor, but also on the dig­i­tal post-pro­cess­ing in the camera/drone. This tech­nol­o­gy is nowa­days called “com­pu­ta­tion­al pho­tog­ra­phy” and has been dri­ven in par­tic­u­lar by the smart­phone industry.

It’s quite impres­sive what mod­ern smart­phones cur­rent­ly offer in this area. For a more in-depth intro­duc­tion to the sub­ject, I rec­om­mend the series of arti­cles on com­pu­ta­tion­al pho­tog­ra­phy on dpRe­view.

There­fore, I was of course very curi­ous to see what the DJI Mini 4 Pro has to offer. As with my pre­vi­ous tests, I again took com­par­i­son shots out of my win­dow. This time I even test­ed 5 cameras:

• DJI Mini 4 Pro
• DJI Mini 3 Pro
• DJI Air 3
• Canon EOS R5 with 24-105mm f/4L IS
• IPhone 15 Pro

Again, I took the images in quick suc­ces­sion. I import­ed them into Adobe Light­room Clas­sic 12.5 and edit­ed them accord­ing to my usu­al work­flow. All of the DJI drone shots were tak­en with the same expo­sure set­tings and processed iden­ti­cal­ly in Light­room. Here are my devel­op­ment set­tings in Light­room Clas­sic 12.5:

I then processed the two shots tak­en with the iPhone 15 Pro and my full-frame Canon EOS R5 in Light­room to make the col­or tone and expo­sure appear rough­ly com­pa­ra­ble to the drone shots. Here you can see the com­plete ver­sion of the sin­gle images cropped to 2048 pix­el horizontally:

In the total view, scaled down to about 3 megapix­els, the shots are not much dif­fer­ent. What some of you may notice is that the fram­ing of the DJI Mini 4 Pro and DJI Air 3 is a bit nar­row­er than that of the Mini 3 Pro, despite the lens­es shar­ing the same focal length. This is because both include a lens pro­file in the DNG files that Light­room already applies automatically:

The DNG file of the Mini 3 Pro does not con­tain such a profile.

Mini 4 Pro vs Mini 3 Pro

Most of you will be inter­est­ed in the com­par­i­son between Mini 4 Pro and Mini 3 Pro. For this pur­pose, here are again some crops tak­en from the Light­room Clas­sic com­par­i­son view in 400%.

First, a com­par­i­son from the cen­ter of the image. Here and in the oth­er com­par­isons, the Mini 4 Pro is shown on the left side:

As in my com­par­i­son of the Mini 3 Pro with the DJI Air 3, the image of the DJI Mini 3 Pro seems sharp­er at the first glance. How­ev­er, a clos­er look reveals that the Mini 3 Pro has been sig­nif­i­cant­ly sharp­ened inter­nal­ly, which results in notice­able sharp­en­ing arti­facts in the form of dou­ble con­tours, espe­cial­ly in the gaps between the bricks.

To demon­strate that the seem­ing­ly increased sharp­ness of the DJI Mini 3 Pro com­pared to the Mini 4 Pro is actu­al­ly just due to more re-sharp­en­ing and not bet­ter optics, I did a com­pa­ra­ble amount of re-sharp­en­ing on the Mini 4 Pro shot in Light­room Clas­sic once:

Post-processed in this way, the sub­jec­tive impres­sion of sharp­ness of the image details of both DJI Mini drones becomes com­pa­ra­ble again.

Here is anoth­er com­par­i­son of the Mini 4 Pro with the Mini 3 Pro:

Again, the right shots of the Mini 3 Pro seem sharp­er at first glance. But they show again some sharp­en­ing arti­facts in the form of dou­ble con­tours (espe­cial­ly in the win­dow frames) and also some colorfringes.

DJI Mini 4 Pro vs DJI Air 3

Now the com­par­i­son of the above crop region with the DJI Air 3:

Here the dif­fer­ence is small­er, per­son­al­ly I like the mod­er­ate sharp­en­ing of the Air 3 quite well, but a com­pa­ra­ble effect can be achieved with a min­i­mal addi­tion of sharp­en­ing in Light­room on the Mini 4 Pro (left) too:

Over­all, I am quite sat­is­fied with both results.

Mini 4 Pro or Mini 3 Pro?

In sum­ma­ry, the images of the Mini 3 Pro appear sharp­er than those of the Mini 4 Pro at first glance, but this is in my opin­ion due to a too inten­sive post-sharp­en­ing of the Mini 3, which cre­ates sev­er­al arti­facts. Per­son­al­ly, I pre­fer the image of the Mini 4 Pro, where I can bet­ter con­trol the degree of re-sharp­en­ing by myself. Unfor­tu­nate­ly, the Mini 4 Pro is not (yet) sup­port­ed by my ref­er­ence RAW devel­op­er, DxO Deep­Prime. With the Mini 3 Pro, I was still able to achieve a qual­i­ty improve­ment with it.

How­ev­er, the dif­fer­ences between the two mini mod­els are not very big. They are only notice­able in strong mag­ni­fi­ca­tion. And I par­tic­u­lar­ly point out once again that the image details shown here were real­ly mas­sive­ly enlarged in order to make the rel­a­tive small dif­fer­ences vis­i­ble at all. So that you can under­stand it bet­ter: If you would enlarge the whole image in the same scale as the details shown here, it would be print­ed out at a usu­al screen res­o­lu­tion of 100 dpi with a width of about 8 meters!

How­ev­er, the dif­fer­ence should hard­ly be notice­able when viewed nor­mal­ly and in usu­al print formats.

So, to put it blunt­ly, when it comes to pho­to qual­i­ty alone, you can still be very sat­is­fied with the Mini 3 Pro.

Looking beyond

Anoth­er very excit­ing aspect is the com­par­i­son with the oth­er two test can­di­dates. Like the DJI drones test­ed here, the actu­al iPhone 15 Pro is equipped with a 48-megapix­el Quad Bay­er sen­sor of a sim­i­lar size (1/1.28″). So what can the cer­tain­ly much more pow­er­ful post-pro­cess­ing of the iPhone make of it?

The result real­ly amazed me. Here are the two above crops for com­par­i­son, again the DJI Mini 4 Pro on the left, the iPhone 15 Pro on the right:

The iPhone demon­strates where “com­pu­ta­tion­al pho­tog­ra­phy” is going. Despite the same sen­sor size and res­o­lu­tion, the iPhone’s image is much more detailed and clear. The iPhone achieves this by tak­ing many pho­tos with dif­fer­ent para­me­ters in quick suc­ces­sion with each shut­ter release and intel­li­gent­ly com­bin­ing the indi­vid­ual images. The result is tru­ly amazing.

This comes very close to my per­son­al ref­er­ence cam­era, a Canon EOS R5 (full-frame sys­tem cam­era with 45 megapix­el res­o­lu­tion) - at least in decent light­ing. Here is a com­par­i­son between Apple iPhone 15 pro (left) and Canon EOS R5 with the stan­dard zoom 24-105 f/4L at 24mm (right):

Résumé

The Mini 4 Pro is a sig­nif­i­cant evo­lu­tion of the Mini 3 Pro in terms of its fea­ture set, omni­di­rec­tion­al obsta­cle detec­tion, flight dynam­ics, range of con­trol, and also in its video capa­bil­i­ties thanks to 4k 60P in all modes and D-LOG-M.

I per­son­al­ly con­sid­er the pho­to qual­i­ty in the 48-megapix­el mode (which I use exclu­sive­ly) to be slight­ly bet­ter thanks to the low­er sharp­ness arti­facts. Over­all, how­ev­er, the dif­fer­ence to the pre­de­ces­sor is not very dra­mat­ic. How­ev­er, in my opin­ion, the improve­ment is not suf­fi­cient as the sole rea­son to switch.

I also appre­ci­ate the much bet­ter inte­gra­tion of the 48-megapix­el mode, which now allows expo­sure brack­et­ing and con­tin­u­ous shoot­ing. Hope­ful­ly, DJI will increase the num­ber of con­tin­u­ous and HDR shots from the cur­rent three to five in a future firmware update, as it did with the Air 3.

The com­par­i­son with the cur­rent iPhone 15 Pro proves that there is still room for improve­ment even with the same sen­sor size due to sheer pro­cess­ing pow­er. The image qual­i­ty - at least in good light­ing con­di­tions - is sur­pris­ing­ly close to that of a sys­tem camera.