ComfyUI Node: ◎ Radiance 32-bit Denoise

◎ Radiance 32-bit Denoise:

RadianceDenoise is a specialized node designed to enhance the quality of images by reducing noise while preserving important details. This node is particularly beneficial for processing high dynamic range (HDR) images, where noise can often obscure fine details and degrade image quality. By employing advanced denoising techniques, such as bilateral filtering, RadianceDenoise effectively smooths out noise in images with varying channel configurations, including RGB, RGBA, and grayscale. The node intelligently handles different image formats, ensuring that the denoising process maintains the integrity of the original image's color and detail. This makes it an essential tool for AI artists and designers who require clean and high-quality images for their creative projects.

◎ Radiance 32-bit Denoise Input Parameters:

d

The parameter d represents the diameter of each pixel neighborhood used during filtering. It determines the extent of the area around each pixel that will be considered when applying the bilateral filter. A larger value of d results in more smoothing, which can effectively reduce noise but may also blur fine details. Conversely, a smaller value preserves more detail but may not remove as much noise. The choice of d should be based on the desired balance between noise reduction and detail preservation.

sigmaColor

sigmaColor is a parameter that controls the filter's sensitivity to color differences. It defines how much influence the color of neighboring pixels has on the filtering process. A higher sigmaColor value allows for greater color variation within the neighborhood, leading to more aggressive noise reduction. However, this can also result in color bleeding, where distinct colors blend together. A lower value preserves color boundaries but may not reduce noise as effectively. Adjusting sigmaColor helps in achieving the desired level of color detail retention.

sigmaSpace

The sigmaSpace parameter determines the spatial extent of the filter's influence. It defines how far the filter considers neighboring pixels in terms of their spatial distance. A larger sigmaSpace value results in a broader area of influence, which can smooth out larger noise patterns but may also blur edges. A smaller value confines the filter's effect to a more localized area, preserving edges but potentially leaving some noise. Balancing sigmaSpace is crucial for maintaining image sharpness while reducing noise.

◎ Radiance 32-bit Denoise Output Parameters:

denoised_frame

The denoised_frame is the output parameter that provides the processed image after the denoising operation. This output retains the original image's dimensions and channel configuration, ensuring compatibility with subsequent processing steps. The denoised frame exhibits reduced noise levels, resulting in a cleaner and more visually appealing image. This output is particularly valuable for enhancing the quality of images used in artistic and creative applications, where clarity and detail are paramount.

◎ Radiance 32-bit Denoise Usage Tips:

• Experiment with different values of d, sigmaColor, and sigmaSpace to find the optimal balance between noise reduction and detail preservation for your specific image.
• For images with significant color variation, consider using a higher sigmaColor to ensure effective noise reduction without losing important color details.
• If your image contains fine details that need to be preserved, start with a smaller d and sigmaSpace and gradually increase them until you achieve the desired level of noise reduction.

◎ Radiance 32-bit Denoise Common Errors and Solutions:

Unsupported channel count, return original

• Explanation: This error occurs when the input image has a channel configuration that is not supported by the node, such as an unusual number of channels.
• Solution: Ensure that the input image is in a standard format, such as RGB, RGBA, or grayscale, before processing it with the RadianceDenoise node.

OpenCV requires contiguous array

• Explanation: This error arises when the input image array is not contiguous in memory, which is a requirement for certain OpenCV operations.
• Solution: Use the np.ascontiguousarray() function to convert the input image array into a contiguous format before passing it to the node.