Unable to add blur filter with max matrix size

[low-batt]

System and IINA version:

• macOS 13.2.1
• IINA 1.3.1

Expected behavior:
The matrix size slider for a blur filter only permits setting valid values.

Actual behavior:
Attempts to add a Blur filter with the Matrix size slider knob moved all the way to the right:

Fail reporting Error occured when setting filters. Please check our parameter format:

IINA log:

18:52:25.623 [player0][d] Adding video filter lavfi=[unsharp=23:23:-0.0:23:23:-0.0]...
18:52:25.623 [iina][w] mpv log: [ffmpeg] error: Parsed_unsharp_0: luma or chroma matrix size too big
18:52:25.623 [iina][w] mpv log: [ffmpeg] error: AVFilterGraph: Error initializing filter 'unsharp' with args '23:23:-0.0:23:23:-0.0'
18:52:25.623 [iina][w] mpv log: [lavfi] fatal: parsing the filter graph failed
18:52:25.624 [player0][d] Failed
18:52:25.624 [iina][w] mpv log: [user_filter_wrapper] error: Creating filter 'lavfi' failed.

mpv log:

18:52:25.623 [iina][w] mpv log: [ffmpeg] error: Parsed_unsharp_0: luma or chroma matrix size too big
18:52:25.623 [iina][w] mpv log: [ffmpeg] error: AVFilterGraph: Error initializing filter 'unsharp' with args '23:23:-0.0:23:23:-0.0'
18:52:25.623 [iina][w] mpv log: [lavfi] fatal: parsing the filter graph failed
18:52:25.624 [player0][d] Failed
18:52:25.624 [iina][w] mpv log: [user_filter_wrapper] error: Creating filter 'lavfi' failed.

Steps to reproduce:

• Play a video
• Click Video filters… under the Video menu
• In the Video Filters window click +
• In the pop-up click Blur
• Slide the Matrix size slider knob all the way to the right
• Click the Add button

• MPV does not have this problem.

This is an IINA problem.

How often does this happen?
Every time.

[low-batt]

Analysis

The FFmpeg unsharp filter documentation indicates 23 is the maximum value permitted:

luma_msize_x, lx
Set the luma matrix horizontal size. It must be an odd integer between 3 and 23. The default value is 5.

luma_msize_y, ly
Set the luma matrix vertical size. It must be an odd integer between 3 and 23. The default value is 5.

IINA is enforcing the minimum and maximum values:

// blur
FilterPreset("blur", params: [
  "amount": PM.float(min: 0, max: 1.5),
  "msize": PM.int(min: 3, max: 23, step: 2, defaultValue: 5)
]) { instance in
   return MPVFilter.unsharp(amount: -instance.value(for: "amount").floatValue,
                           msize: instance.value(for: "msize").intValue)
},

The FFmpeg code generating the error looks like:

if (s->luma.scalebits >= 26 || s->chroma.scalebits >= 26 || s->alpha.scalebits >= 26) {
    av_log(ctx, AV_LOG_ERROR, "luma or chroma or alpha matrix size too big\n");
    return AVERROR(EINVAL);
}

The error message is slightly different because FFmpeg has added an alpha parameter to this filter.

The code calculating scalebits looks like:

fp->steps_x = msize_x / 2;
fp->steps_y = msize_y / 2;
fp->scalebits = (fp->steps_x + fp->steps_y) * 2;

The FFmpeg documentation does not mention that if both the horizontal and the vertical size are set to their maximum values this restriction on the combined size will be exceeded.

[low-batt]

IINA 1.3.2 contains the fix for this issue.