Test Images for Sony ARW2 Stripe Defect

To provide test images for KARWY-SR, I've been trying various ways to induce these stripe artifacts. Here are some of the most successfully-induced (i.e., worst looking) artifacted images I've been able to produce. I intend to continue to grow the set of test images posted here as new examples of artifacted images are produced. All images linked on this page may be freely used by others seeking to document, understand, or repair this artifacting, but should be cited as the work of Hank Dietz at the University of Kentucky.

For each image, three JPEGs are shown and the first two have raw files linked. The first raw image is the compressed ARW2 produced by the camera. The second raw image is the compressed ARW2 produced by KARWY-SR. The JPEGs were created using identical processing with RawTherapee to create interpolated TIFF images that were loaded into GIMP and differenced. The non-OOC JPEGs were created using GIMP, with the auto-leveled difference as the third image.

Note that the difference image only shows what was detected and repaired by KARWY-SR, which might not be "everything" that was wrong. Some of the difference images also show blocking artifacts; these are not related to the stripes, but rather come from the lossy compression used in ARW2 files being interpreted slightly differently when RawTherapee applies its default processing to the original and repaired ARW2 files.

Most of the shots were taken using bright lights and angles intended to cause lens flare, but it is worth noting that it can be very difficult to get any artifacting that way. It seems the light needs to hit at very specific angles, which a diffuse bright spot sometimes hits, but more often doesn't. On the cameras that are less prone, the only reliable way to induce the patterns was to shine an ultra-bright LED flashlight across the bare open lens mount. This revealed that some cameras produce stripes that are darker, not lighter. For each camera, we summarize the strength and type of striping and how commonly one might see it in high-risk scenes shot using a lens.

Sony RX-100 IV (no stripes)

This seems to be a very similar camera to the RX-100 V, but no stripes. Well, the sensor doesn't have the 315-point phase detection that the RX100 V has either, just contrast detection, ... so I guess that's enough explanation.

Sony RX-100 V (moderate bright stripes, common)

This was shot against a dark background using the halogen desk lamp, seen to the upper right of the scene, as the sole light source. First, here's the image as handled by 20180315 alpha test version of KARWY-SR:

and here's the same thing via KARWY-SR 20180318 beta:

Note that any color changes are due to slightly different settings in RawTherapee and/or GIMP.

Sony NEX-5 (no stripes)

No artifacting visible, even with no lens mounted.

Sony NEX-7 (no stripes)

No artifacting visible in any tests thus far -- which includes the exact same scene that caused the artifacting in the A6500 shot below (DSC03154) and even two more attempts using the Samyang 85mm f/1.4 and halogen light. It's difficult to be sure, but this camera seems immune... which would make sense in that it doesn't have PDAF sensels.

Sony A6000 (heavy bright stripes, common)

Before I was able to test mine, there is a photo of a cat (DSC02954) somebody else took with an A6000 that has terrible artifacting. Interestingly, KARWY-SR 20180315 does not complain, but makes no repairs to that file whatsoever. This was the main motivation for version 20180318, which repairs that example well. Here's one I shot using a Samyang 85mm f/1.4:

Sony A6500/ILCE-6500 (minor bright stripes, common)

This is essentially a shot of nothing but my dark basement, with heavy flare from the halogen desk lamp seen at the extreme right edge. The lens used was a Samyang 85mm f/1.4 on a Sony LA-EA4 adapter; this lens and adapter were selected because they produce a full-frame image circle, thus allowing a larger than normal amount of light to be coming from unusual angles between the lens and the sensor. The lens also has a very large rear element, also increasing the range of ray angles that could enter the camera body. Here's the 20180315 version:

and here's the 20180318 version:

Here's an even more dramatic version shot with no lens mounted:

Sony A7/ILCE-7 (minor dark stripes, very uncommon)

No artifacting observed in two dozen attempts... then I tried without a lens mounted. The artifacts were mildly visible, but strictly as darker red pixels.

Sony A7II/ILCE-7M2 (moderate dark stripes, very uncommon)

No artifacting observed in two dozen attempts... then I tried without a lens mounted. The artifacts were moderately visible, as darker red and green pixels.

Sony A7RII/ILCE-7RM2 (heavy bright stripes, very uncommon)

It took a lot more tries to get artifacting from this camera... but I did eventually get them. Here's again with the Samyang 85mm f/1.4 and a halogen lamp. Interestingly, although this shot shows very clear artifacts, none of the others shot with just slightly different framing have even a trace of artifacts.

The Aggregate The only thing set in stone is our name.