Suppress Image Noise

Noise is very common in images taken by both digital cameras and film cameras. Even high-end professional cameras can produce images with a high level of noise at low light or high ISO settings. Topaz Vivacity™ offers three filters, Topaz Clean™, Topaz Clean (YCbCr)™, and Topaz DeNoise/EQ™ for cleaning up noise from images. Topaz Clean™ and Topaz Clean (YCbCr)™ use a unique structural enhancing filter that remove noise and increase the crispness of edges at the same time. Topaz DeNoise/EQ™ is a noise filter based on wavelet transformation. It tends to create smooth looking images. Which filters to use depends on the image content and your intention.

Use Topaz Clean™

Here is a high resolution image taken by an OLYMPUS SLR digital camera at 3264x2448 resolution and ISO1600. It’s a nice picture but strong noise is visible upon detailed examination. The first figure in the next page is a small portion of this same image viewed at original resolution. The noise caused by high ISO is obvious. This picture is essential unusable if you want to print it out at 8x11” or larger, Apply Topaz Clean™ with the following parameter settings, you get the image with much less noise and clear edges.

Topaz Clean™ is designed to clean up noise from RGB images. Unlike most modem noise removal techniques that mostly based-on wavelet approach, Topaz Clean™’s unique approach allows you to achieve excellent results without calibration camera to create “noise profile”, or fiddling with a large number of parameters. Moreover, it creates crisp looking edges and reduces color bleeding.

There are only two key parameters, “Threshold(G)” and “Clean Radius”. “Threshold” sets a level under which small random features are considered to be noise and tend to be filtered out. The “Clean Radius” specifies the radius of a circle inside which pixels affects the cleaning result. For noisier images, use higher “Threshold” and “Clean Radius”. However, do not use a “Threshold” value higher than necessary, as this could result in useful image details being suppressed.

You can apply different thresholds on R, G, and B channels. Use “R Threshold vs Main” and “B Threshold vs. Main” to adjust the relative values of the threshold for the red (R) channel and blue (B) channel. In our example, “R Threshold vs Main” is 1.3 since the red channel has much more noise than the green one, which is the main channel. Setting this value to 1.3 allows more suppression of red channel noise, resulting in a better result. To examine noise levels in each channel, use the “Channel” tab page in Photoshop to view each channel separately. Other hosts have similar methods.

Topaz Clean™ also contains a set of parameters that sharpens the image at the same time. These parameters are “Sharpness”, “Sharpness Radius”, and “Line Accent”. They are basically a simplified version of Topaz Sharpen™ integrated here for your convenience. Detailed explanations of each can be found in the section Sharpen Images.

“Iterations” specifies how many times the filtering be repeated. Applying Topaz Clean™ more than one times can achieve results that are not achievable with one iteration. In the images on the next page, 4 iterations are used to have this “overly clean” effect:

Topaz Clean(YCbCr)™ performs similarly to Topaz Clean™. The difference is that Topaz Clean(YCbCr)™ works internally in YCbYr color space. Both Topaz Clean™ and Clean(YCbCr)™ can work in gray image mode. Therefore, it can be used to process the image in any mode, channel by channel.