Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Several small improvements #3

Open
wants to merge 16 commits into
base: master
Choose a base branch
from
Open

Several small improvements #3

Changes from 1 commit
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
05c4587
Comma separated file for Gnuplot
Jan 12, 2020
0a01bcd
Add variables for grid size
Jan 12, 2020
ac42708
Averaging over the entire grid cell
Jan 12, 2020
a889235
Read sensor model from the RAW header and set appropriate black level
Jan 12, 2020
0f526c7
Lens shading table binary output
Jan 12, 2020
423977f
Search the brightest area in the image
hasepompase Jan 21, 2020
c0f3877
fix indentation
Jan 21, 2020
b2a950e
Changed calculation of grid_width and grid_height to initial method
Jan 21, 2020
f290689
Removed double calculation loop, added parameters for cell size
Jan 23, 2020
22f3acf
Add getop parameter parsing and help text output
Feb 7, 2020
5252078
Add output format parameter
Feb 7, 2020
1213f9e
Fix format
Feb 7, 2020
18d809e
Fix format
Feb 9, 2020
b239de6
Added support for HQ cam
Jun 24, 2021
8746d6a
Revert "Added support for HQ cam"
Jun 25, 2021
83d9275
Added support for HQ cam (with proper indentation)
Jun 25, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
45 changes: 21 additions & 24 deletions lens_shading_analyse.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -310,35 +310,32 @@ int main(int argc, char *argv[])

fprintf(header, "//%s - Ch %d\n", channel_comments[i], channel_ordering[bayer_order][i]);
uint16_t *line;
for (y=16; y<single_channel_height+32; y+=32) //Grid size is 64x64, but the component tables are subsampled due to the Bayer pattern
for (y=0; y<grid_height; y++)
{
if (y>=single_channel_height)
line = &channel[(single_channel_height-1)*(single_channel_width)];
else
line = &channel[y*(single_channel_width)];

for(x=16; x<single_channel_width; x+=32)
int y_start = y*32;
int y_stop = y_start + 32 <= single_channel_height ? y_start + 32 : single_channel_height;
for (x=0; x<grid_width; x++)
{
//Average 3 pixels horizontally to give some noise rejection
int avg = line[x] + line[x-1] + line[x+1];
int gain = (middle_val*3) / avg;
int x_start = x*32;
int x_stop = x_start + 32 <= single_channel_width ? x_start + 32 : single_channel_width;

uint32_t block_sum = 0;
uint16_t block_px = 0;

for(int y_px = y_start; y_px < y_stop; y_px++){
line = &channel[y_px*(single_channel_width)];
for(int x_px = x_start; x_px < x_stop; x_px++){
block_sum += line[x_px];
block_px++;
}
}
int gain = (middle_val*block_px) / block_sum;
if (gain > 255)
gain = 255; //Clip as uint8_t
gain = 255; //Clip as uint8_t
else if (gain < 32)
gain = 32; //Clip at x1.0
gain = 32; //Clip at x1.0
fprintf(header, "%d, ", gain );
fprintf(table, "%d %d %d %d\n", x, y, gain, i );
}
//Compute edge value from the very edge 2 pixels.
{
int avg = line[single_channel_width-2] + line[single_channel_width-1];
int gain = (middle_val*2) / avg;
if (gain > 255)
gain = 255; //Clip as uint8_t
else if (gain < 32)
gain = 32; //Clip at x1.0
fprintf(header, "%d,\n", gain );
fprintf(table, "%d %d %d %d\n", x, y, gain, i );
fprintf(table, "%d %d %d %d\n", x * 32 + 16, y * 32 + 16, gain, i );
Copy link
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This feels wrong.
The ISP interpolates the values at the centre of each grid cell, and interpolates between them. This is averaging over the entire cell, therefore is just a smoothing filter.

Enlarging the average to 5x5, or even up as far as 9x9 would be reasonable, but otherwise you're ignoring a large amount of the detail.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I do not know which algorithm is used for the interpolation. I don't see a problem with this approach. From my point of view and for my applications the results are better.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I've been thinking again. If you look at lens shading across the width (alternatively across the height), the brightness will be a function similar to figure a. The average value changes only slightly with the width of the viewed area. However, the brightness signal from the sensor is subject to noise (figure b). A wider area improves the average value. Only in case of peaks smaller areas are better (figure c). In this case, however, the peak would have to lie exactly in a calculation area.

lens_shading

Or am I misunderstanding something?

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Integrated a parameter for the cell size.

}
}

Expand Down