robertgawron / supper-resolution Goto Github PK

Some sources (unfortunately links stay on my broken computer) indicates that other function that PSF may be used in SR algorithm. PSF is stored as array, it can be array multiplied by scalar, or invertible PSF matrix. Could this be use here to obtain better results? My previous tests gave miserable results, but it's worth to check once again.

Fix mkdirOutput() in SampleCreator.py according to:
http://stackoverflow.com/a/5032238

Traceback (most recent call last):
File "SRRestorer.py", line 153, in
origSizeX = samples[0][1].size[1] * scale
IndexError: list index out of range

It's extremely painful to install numpy on Windows 7 and in addition API of this library is hard to use. A C library should be used instead.

It's needed to be able to start checking results of SR implementation on real word pictures.

IIRC some people made it by scanning dot marked on paper, then scanned image of this dot was representation of PSF. It looks simple, but could this be use also for cameras?

Framework: unittests
https://docs.python.org/3.4/library/unittest.html

Switch to C library. SRImage should take care of the interface stuff, so no need to change implementation of other components. Probably Imagemagic (MagickCore API) should be used.

This is part of moving project out to a C library and providing only bindings for other languages.

python3 ./SRRestorer.py /Users/pro16/Dropbox/Terminal/supper-resolution-master/input/
Estimate Motion Between Sample And Original Image
Restore SR Image
Creating Camera Model
Size Of Estimated Original: 420x520
 0: estimation error: 14.838214
 1: estimation error: 14.335206
 2: estimation error: 13.815206

Every time after running I have the same problem, please add some more information for this error describing.

Source image in /input folder is this.

I'm quite new to image processing and I am doing a project based on super-resolution and was tinkering with your code.
I noticed that it fails with a code when using png files instead of tiff files.
Camera.py", line 48, in take
odata[:,:,0] = self.doOffset(odata[:,:,0], offset)
IndexError: too many indices for array
is there any formatting difference between the two?

Add a log to SRRestorer - ErrorEstimation.gnuplot should read its data from it.

Add entry on wiki how to use it.

can you show what's like after supper resolution from you example image?
@RobertGawron
It looks like you don't init the image by averaging LR images but paper does.
Any suggestion?

Doc is missing, this makes code harder to modify.

1. Split to smaller methods.
2. Analyze and remove magic numbers.

To make the project more user friendly, a default directory name for input sample should be selected if it wasn't passed as an argument from command line.

This was already done for SampleCreator, should be done in similar way for SRRestorer.