joanvr / reconstructor Goto Github PK

Our project is divided in three applications:
	
	CSCLI: command line version of the reconstructor
	CSGUI: graphic interface version of the reconstructor
	CSGEN: utility to create holes to ppm files

The usage of the programs is:

./CSGEN inputName sampleRatio
	inputName: image file to erase pixels from
	sampleRatio: how many pixels will be kept

	The output has two files: 
	  - the image with some pixels missing
	  - the decision matrix for this image
	
	Example:
		./CSGEN lena.ppm 0.6  
		Output: 	lenaH60.ppm 	lenaH60.txt
	
./CSCLI inputFile sigmaLambda blockSize transform [rgb/yuv] [nthreads]
	inputFile: name of the ppm image file. We assume that the decision
			   matrix is located in the same directory and has the same
			   name but with .txt extension.
	sigmaLambda: tolerance measure. A double, typically less than 1.0
	blockSize: 16 or 32
	transform:  0 - random ones
				1 - random sqrt(3)
				2 - discrete cosine
				3 - daubechies 8
				4 - cdf 9/7
	rgb/yuv: optional. Assumed rgb as default.
	nthreads: optinal. Assumed 4 threads as default.

	The output has three files:
	    0-original.ppm
			the read image with missing data (for test purposes)
		1-transformed.ppm
			the transformed image that we obtain as Gurobi's solution.
		2-reconstructed.ppm
			the recovered good image, the inverse transform of the previous.

./CSGUI
	It has no inputs, as all parameters are set through a Qt interface.
	The output is the same as CLI version.