I am implementing lz77 compression for images. But if I insert an RGB image the output is wrong could any of you help me?

Question

when i insert a grayscale image then the algorithm works if i insert a rgb image the output is wrong. is the way I work with the rgb image correct? I don't understand why can someone help me?

library used

from PIL import Image

import numpy as np from cv2 import cv2

 def lz77Compress (image,sw,lab):
    img = cv2.imread(image)
    print("Initial size", img)
    flat = np.array(img).flatten()
    #cv2.imshow("input", img)
    row = img.shape[0]
    col = img.shape[1]
    tot = row * col
    slidingWindows = sw
    lookAhead = lab 

array of tuyple and charactert

encodedTuple = np.array([]) 
encodedChar = np.array([])

pointer in the search buffer

sbPointer = 0
while sbSize + sbPointer < tot :
    max_match = 0
    max_match_go_back = 0        
    selChar = sbPointer + sbSize

empty sequence

   seqY = np.array([])
    for i in range(sbPointer,sbPointer + sbSize):
        if(flat[i] == encodeCharacters):
            seqY = np.append(seqY,i)

check if there is a match in the lookAheadBuffer

 if(seqY.size == 0 ):
        encodedTuple = np.append(encodedTuple,(0,0))
        encodedChar = np.append (encodedChar,encodeCharacters)
    else:

        for j in seqY:
        #lunghezza della corrispodenza
            matchLenght= 0
            returnBack = selChar -i
            it = 0 
            while selChar + it < tot :
                if flat[it + j] == flat[selChar + it]:
                    matchLenght +=1
                    it +=1
              # se non trova corrispondenze
                else: 
                    break
            if matchLenght>max_match:
               max_match = matchLenght
               returnBack = max_match_go_back

save corrispondence and tuple in array

encodedTuple = np.append(encodedTuple,(max_match_go_back,max_match))
encodedChar = np.append(encodedChar,flat[selChar + max_match - 1])
        
       sbPointer+= 1 +max_match

**save of encodedTuple, encodedChar , file compresses.txt and imsize of decompression **

print("Prova", encodedTuple, encodedChar)
print("ArrayBin",encodedTuple.tolist(), encodedChar.tolist())
np.save("encodedTuple", encodedTuple) 
np.save("encodedChar", encodedChar)
a = encodedTuple.tolist()
b = encodedChar.tolist()
c= (a,b)
print("File compresso in : Copressed.txt")
output = open("Compressed.txt","w+")
output.write(str(c))
imgSize = open('imgSize.txt', "w")
imgSize.write(str(row) + '\n')  # write row dimension
imgSize.write(str(col) + '\n')  # write col dimension
cv2.waitKey(0)
cv2.destroyAllWindows()

**Main **

path = "./im3.jpg"

lz77Compress (path,500,500)

Answer 1

In most cases, RGB image has 3 color channels, and gray image has 1 color channel.

• Assume img is a NumPy array that represents a gray image:
  img.shape is going to be (rows, cols), and img.size is rows*cols.
• Assume img is a NumPy array that represents a RGB (or BGR) image:
  img.shape is going to be (rows, cols, ch), and img.size is rows*cols*ch.
  In case of RGB, ch = 3.

I downloaded your code from github.
I saw you already corrected the code to support RGB images.

I can't reproduce the error in line 35 of image.py
Did you fix it?

The issues I could find:

• You are not closing the files:
  Close the files at the end of lz77Compress:

   # We must close the files.
   output.close()
   imgSize.close()
  

You better also close imsize after reading row, col, ch in lz77Decompressor.

• When using cv2.imshow("output", decodeArray), you need the type of decodeArray to be uint8.
  You may use the following code to show the output:

   cv2.imshow("output", decodeArray.astype(np.uint8)) # Show as uint8
  

In case you want to support both gray and RGB, you can check if len(img.shape) is 2 or 3, and add some logic at the Decompressor.

Note:
I didn't review the entire code, so there might be problems that I missed.