labkit.image.sharpen

Description

Converts imageIn to clamped RGB double and constructs an unsharp mask by subtracting a square local mean from the image. The detail mask is scaled by 2*amountPct/100 and added to the original. Each RGB channel is treated independently, and the result is clamped to [0,1].

radiusPx selects an odd mean-filter width of 2*round(radius)+1, with a minimum width of 3 pixels. This function uses a box blur rather than a Gaussian blur, so radius describes neighborhood size rather than a Gaussian standard deviation.

Inputs

imageIn

Numeric or logical grayscale, RGB, or multichannel image.

amountPct

Numeric scalar effect strength in percent. Negative values are treated as 0.

radiusPx

Numeric scalar blur radius in pixels. Values below 0.5 are treated as 0.5 before rounding.

Outputs

imageOut

M-by-N-by-3 double image with values in [0,1].

Failure Behavior

Unsupported image classes or channel shapes propagate errors from im2double or ensureRgb. amountPct and radiusPx must be numeric scalars; incompatible values raise the originating conversion or indexing error.

Example

imageIn = zeros(9);
imageIn(3:7, 3:7) = 0.6;
imageOut = labkit.image.sharpen(imageIn, 50, 1);