Create binary image from compressed domain values in text file

These are the .txt and .csv files of one of the frame. I want to be able to create a binary image from this file, I have the index of the block, the width, height, x position, y position and total number of bits. I wnat to use total no.of bits as pixel intensity and create an image based on that. Convert that no.of bits column to 0s and 1s either by didving by the largest value or taking the logarithmic value and create the image. Please guide me on how I should proceed further :)

clear all; 
clc

cutable = readtable ('frame1.csv'); % i,x,y,w,h,no.of bits
carray = table2array (cutable);


len = length (carray);
idx = carray (:,1);
w = carray (:,2);
h = carray (:,3);
x = carray (:,4);
y = carray (:,5);
b = carray (:,6);

Answer 1

[Walter Roberson]

cutable = readtable ('frame1.csv'); % i,x,y,w,h,no.of bits
carray = table2array (cutable);
len = length (carray);
idx = carray (:,1);
w = carray (:,2);
h = carray (:,3);
x = carray (:,4) + 1;
y = carray (:,5) + 1;
b = carray (:,6);

maxx = max(x + w - 1);
maxy = max(y + h - 1);
img = zeros(maxy, maxx);

for K = 1 : numel(x)
    tx = x(K); ty = y(K);
    img(ty:ty+h(K)-1, tx:tx+w(K)-1) = b(K);
end

img8 = uint8(rescale(img, 0, 255));

subplot(2,1,1);
image(img8)
colormap(gray(256))
title('linear');

minb = min(b);
scaleb = log(b ./ minb);

img = zeros(maxy, maxx);

for K = 1 : numel(x)
    tx = x(K); ty = y(K);
    img(ty:ty+h(K)-1, tx:tx+w(K)-1) = scaleb(K);
end

subplot(2,1,2);
imgL = uint8(rescale(img, 0, 255));
image(imgL)
colormap(gray(256))
title('log')

It looks to me as if the x, y indexing is not correct

You indicate that the first column is the index, but you have multiple entries with the same index, such as

64,32,32,32,0,449
64,16,16,32,0,139

It looks to me as if the first four columns have to do with position and size. However you have entries such as

0,64,64,0,0,2316

and if we try to interpret those as x y w h or w h x y then we seem to encounter either x or h that are 0, which does not sound correct.

[Yousra Ashfaq]

Thank you @Walter Roberson for your answer. :) I was able to create a feature image from your answer here: https://www.mathworks.com/matlabcentral/answers/1780080-create-an-image-from-x-and-y-locations-with-greyscale-value?s_tid=srchtitle

I have attached the original image and feature image attached in my answer. The csv file I used is also attached. However I'd like to fine tune the feature image now to show obejcts in the image more clearly by using log or thresholding or any other method. Do let me know if you have any suggesrions :) I used the 4th, 5th and 6th columns as x, y and z respectively.

[Image Analyst]

You aren't trying to track individual jockeys or horses with images like that are you? Because that would be nearly impossible. But start here:

https://www.mathworks.com/help/vision/ref/maskrcnn.segmentobjects.html?s_tid=srchtitle_segmentobjects_1

[Yousra Ashfaq]

@Image Analyst I am working in video compressed domain. I am experimenting on both hevc and vvc. I have extracted features from their reference softwares and using those features I am trying to detect either background/foreground or moving objects. The number of bits extracted are high when there's movement and low when there's not much movement, thats why we were able to create an image using x, y coordinates and number of bits :)

If you have any suggestion on how to proceed with my problem, let me know :)