Sharpening filter

As explained in Deep Learning with JAX

A filter designed to increase local contrast and make image details appear crisper.

Want to recover detail after denoising? A sharpening filter increases contrast between each pixel and its neighborhood. Its kernel places a large positive value at the center and negative values around it, then normalizes the values so filtering stays within a useful range. Clipping remains a final safeguard for values outside the allowed range. section 3.1.4

Sharpening a blurred image

A 5×5 kernel uses 50 at the center and −1 elsewhere. After division by the kernel sum, apply it with the same color-channel convolution used for blurring: img_restored = color_convolution(img_blur, kernel_sharpen). The result restores some sharpness while retaining the earlier noise reduction. section 3.1.4

Figure 3.8 Original image (top left), noised (top right), denoised but blurred (bottom right), and sharpened (bottom left) versions of the image

Compare the original, noisy, blurred, and sharpened image versions to see the tradeoff between denoising and recovered sharpness. section 3.1.4

How Deep Learning with JAX covers it

How deep it goes

Explains kernel structure, normalization, channel-wise convolution, and clipping, but does not derive filter design mathematically.

Examples

Uses a normalized 5×5 kernel and compares original, noisy, blurred, denoised, and sharpened images.

What it assumes

Assumes image arrays and convolution-based filtering, while demonstrating channel handling and output clipping.

What you can do after

Apply it via channel-wise convolution to restore sharpness in a blurred image.

How it connects

Connects sharpening filters with convolution, image filtering, denoising, kernels, and local contrast.

liveBookmanning.com

Read Deep Learning with JAX on liveBook Get the book at manning.com

Related concepts

• batch normalizationused by · Batch normalization is a neural-network technique that normalizes activations using statistics computed over a mini-batch, with learned scale and shift parameters and typically running statistics for inference.
• denoisingaddresses · Denoising is the process of removing unwanted noise or corruption from data while preserving meaningful structure.
• exponential moving averagecontrasts with · An exponential moving average is a recursively updated weighted average that gives greater influence to recent observations while exponentially decreasing the influence of older ones.
• local averagingcontrasts with · Local averaging is a smoothing technique that replaces values with arithmetic means computed from nearby observations within successive or sliding windows.

This explainer was generated by Manning's knowledge graph from the text of Deep Learning with JAX. The section links point to the parts of the book it draws on.