Amazing: 3mp vs. 5mp comparison

[darkness]

I was reading a review of Sony DSC-V1 and the author made a claim of shooting in 3mp on the camera gives a resolution comparable to the native 5mp resolution. Of coz in my mind I tot that sounds really ridiculous, but the author had an image there to prove his point.

Curious, I downloaded the V1's resolution chart from dpreview and did a little experiment. The results are really quite amazing:

1. Load the resolution chart in Photoshop.
2. Resize to 3mp (2048 x 1536) bicubic.
3. Resize back to 5mp (2592 x 1944) bicubic.

Then I took a crop of both the original 5mp picture and the 5mp -> 3mp -> 5mp crop and put them into an animated gif for comparison of the amount of loss in resolution.

The result: the amount of loss in resolution is really quite minimal. This means that shooting in 3mp on the camera may be feasible if you need to save the space on the memory stick (since they are so freaking expensive). Of coz, purists would still want to shoot in 5mp.

Disclaimer: This does not show that 3mp cameras are comparable to 5mp cameras. When you shoot 3mp in a native 5mp camera, the downsizing done in camera produces a higher quality picture than a picture shot in a native 3mp camera. (Eh, can understand my poor wordings?? )

 

[Jed]

Actually, what you're demonstrating is more that 5mp files straight from the camera really aren't that good. Because of Bayer interpolation and other reasons, a straight file from the camera has a long way to go to reaching true 5mp quality. So by downsizing and resizing back up, you're not giving away an awesome amount.

In your test you really should have included a straight 3mp file from the camera as well. Depending on whether your camera shoots 5mp and then downsamples to 3mp (in which case your results will be largely the same), or shooting with only 3m photosites, you might get more divergent results.

 

[darkness]

The author in the review that I quoted above has an example of his 3mp native vs. his 5mp native. The reason why I can't do such a comparison is becoz I don't have the camera!!

But I assume that the downsizing done in camera is as good as the one done in Photoshop. (At least Phil Askey seems to suggest so.) I also tried downloading his "watchface" picture from that page, but his 5mp and 3mp pictures are *slightly* different so a direct crop would not match.

Also did a non-scientific comparison of his 5mp standard compression (1.2mb) vs. his 3mp fine compression (1.4mb) and found that the 3mp fine seems to contain less artifact than taking the 5mp standard and downsizing it to 3mp. Perhaps that extra 200kb really helps... :dunno:

 

[denizenx]

Jed is very right if you get no quality loss when u up it back to 5mp LOL... sounds like the new s50/G5s