Photographic data collection

Data collection with a digital camera

A tailored test sheet was photographed with a digital SLR (single lens reflex) camera.
 Here is a downsampled version:

The idea is to take pictures of exactly the same (above) view with
 (1) Correct focusing and low noise (low noise means small ISO sensitivity 100),
 (2) Incorrect focusing and low noise (ISO 100), and
 (3) Incorrect focusing and high noise (ISO 3200).

We recorded one image of type (1), three images of type (2) and three images of type (3).
 It is important to take all these photos in camera RAW format instead of a format involving
 lossy compression, such as jpeg. In addition to not losing information in compression,
 the RAW files have the advantage that they are saved in 16 bit format instead of 8 bit.
 Namely, 8 bit images allow only 256 different levels in each pixel (and each RGB color component),
 whereas 16 bit files can store the full dynamic range of the analog-to-digital converter of
 the camera, typically 12 or 14 bit (4096 or 16384 levels in each pixel, respectively).

The original images were color photographs containing all RGB channels (red, green, blue).
 We picked out just the red channel and consider the result as a black-and-white, or grayscale, image.

Here are jpeg versions of all the seven images recorded. These are just for viewing purposes;
 uploading all the RAW images is impractical as they are 28 MB in size each.
 (1) Correctly focused low-noise jpeg image
 (2) blur100_1.jpg; blur100_2.jpg; blur100_3.jpg
 (3) blur3200_1.jpg; blur3200_2.jpg; blur3200_3.jpg

From each of the seven images, we extracted the point spread functions of different sizes,
 and portions of the images containing text. Here is the Matlab routine that was used:
read_image_files.m.Unfortunately, the images are not perfectly aligned,
 so some manual registration may be needed for optimal results.

Note that to use any of the Matlab files on this page you need a working directory with
 the .m files, and a subdirectory called "data" where you store all .mat files.

Here are the PSF images from each of the seven images:
 (1) tarkka100PSFdata.mat
 (2) blur100_1_PSFdata.mat; blur100_2_PSFdata.mat; blur100_3_PSFdata.mat
 (3) blur3200_1_PSFdata.mat; blur3200_2_PSFdata.mat; blur3200_3_PSFdata.mat
 Extracting one of the four different-sized PSF functions can be done with this Matlab
 routine: PSF_comp.m(possibly modifying the indices where the PSF is taken).

Here are the text portions of each of the seven images:
 (1) tarkka100text.mat
 (2) blur100_1_text.mat; blur100_2_text.mat;blur100_3_text.mat
 (3) blur3200_1_text.mat; blur3200_2_text.mat; blur3200_3_text.mat

You can compare the measured blurring and simulated blurring by using and modifying
 the following Matlab routine: blur_compare.m