Tools

3x3/4x4 color matrix real-time preview

Bayer RAW to RGB demosaic: bilinear interpolation reconstructs missing color channels

Estimate illuminant color temperature and apply R/B channel gains to remove color cast: Gray World / White Patch

Gamma curve real-time preview

Three spatial denoising methods: Median for salt-and-pepper, Gaussian smoothing, Bilateral edge-preserving

Inter-frame recursive blending temporal denoising — multi-frame averaging to reduce noise, with √N law demonstration

Unsharp Mask sharpening: Gaussian blur extracts high-frequency detail mask, enhances edges and textures, threshold prevents noise amplification

Binning downsampling preview

Custom operator order with real-time stacked preview

RGB ↔ YUV / YCbCr / HSV / XYZ bidirectional conversion with quantization error analysis

RGB/luminance histogram + histogram correction (auto-contrast / equalize / gamma) + clipping detection + luminance statistics summary

3D LUT lookup-based color grading — trilinear interpolation for cinematic/camera color style mapping

HDR→SDR tone mapping — Reinhard / ACES classic curves comparison, recovering highlight details

Mertens-style multi-exposure fusion — simulate bracketed exposures from a single image, weighted fusion to recover highlight and shadow details

8-anchor LUT tone curve adjustment with drag editing and real-time preview

Contrast Limited Adaptive Histogram Equalization (luminance domain), bilinear interpolation between tiles

Subtract sensor dark-current offset to restore the true black level baseline

Detect and fix sensor defective pixels: neighborhood median comparison, replace when deviation exceeds threshold

Correct lens vignetting: apply radial gain based on pixel distance from center to compensate for edge light falloff

Canny multi-stage edge detection: Gaussian smoothing + Sobel gradient + NMS + double threshold hysteresis

Sobel gradient edge detection: 3x3/5x5 kernel convolution with x/y/magnitude directions

Laplacian second-derivative edge detection: 4/8-neighbor kernel, direction-independent, optional Gaussian pre-smoothing

Prewitt/Roberts first-difference edge detection: 3x3 uniform weighting / 2x2 cross-difference

Harris corner detection: structure tensor eigenvalue analysis, distinguishing flat/edge/corner regions

Hough transform line detection: parameter space voting, peak extraction, line segment overlay

Mean blur (Box Filter): 3x3/5x5/7x7 normalized window averaging with separable passes and shift-right normalization

Median filter: window sorting to extract median — classic salt-and-pepper noise removal with superior edge preservation

Noise injection: Gaussian / salt-and-pepper / uniform noise with adjustable intensity for testing denoising algorithms

Adaptive median filter: classical two-stage algorithm that dynamically adjusts window size based on local noise density, outperforming standard median under high-density salt-and-pepper noise

Non-Local Means (NLM): weights each pixel by patch similarity across the search window, preserving textures and repeating structures better than local filters

Guided Filter (He et al. 2010): edge-preserving smoothing via local linear model with box-filter O(1) per-pixel computation, approximates bilateral filter

Morphological erosion: 3x3/5x5 structuring element sliding window minimum, removes isolated white pixels and thin lines

Morphological dilation: 3x3/5x5 structuring element sliding window maximum, fills small holes and bridges gaps

Morphological opening: erosion followed by dilation, removes isolated noise while preserving shape

Morphological closing: dilation followed by erosion, fills small holes and bridges narrow gaps

Morphological gradient: dilation - erosion, extracts edge and contour response

Top-Hat: src - opening(src), extracts bright details and small bright spots

Black-Hat: closing(src) - src, extracts dark details and small dark spots

Image crop (ROI extraction): specify x/y/width/height, auto-clamp out-of-bounds, zero-buffer window addressing

Image resize: nearest neighbor (step counter avoids divider) and bilinear interpolation, FPGA column address remapping

Image mirror flip: horizontal (column reverse) or vertical (row reverse), frame buffer address remapping

Image rotation: 90 / 180 / 270 degree CCW steps (nearest neighbor), width/height swapped at 90 / 270

Affine transform: 2x3 matrix for rotation/scale/shear/translation, parallelism preserved, inverse mapping + bilinear interpolation

Perspective transform: 3x3 homography for perspective effect (vanishing point), lines preserved, homogeneous coordinates + perspective divide

Radial lens distortion correction (Brown-Conrady model): barrel/pincushion visualization, k1/k2 coefficient adjustment, reverse mapping + bilinear interpolation

Integer-factor image upscaling (2x/3x/4x) with three interpolation comparison (nearest/bilinear/bicubic Catmull-Rom) + interpolation quality teaching

Otsu global threshold segmentation: iterates 256 gray levels maximizing between-class variance, auto-finds optimal threshold for one-click binarization

Adaptive threshold: local mean/Gaussian weighted minus C as per-pixel threshold, better than global for uneven illumination

K-means unsupervised clustering segmentation: pixel gray-level feature clustering to k centroids with iterative convergence for region segmentation

Region growing segmentation: seed points + similarity tolerance with 4-connected BFS growth for region-based image segmentation

Watershed segmentation: distance transform + marker extraction + BFS expansion, morphology-based image segmentation with marker-controlled approach to prevent over-segmentation

ITU-R BT.601 luminance grayscale: Y = 0.299R + 0.587G + 0.114B

Linear contrast stretch: maps gray range to [0, 255], enhances low-contrast images

Fixed threshold binarization: gray > threshold -> white (255), else -> black (0)

Single-Scale Retinex (SSR): log-domain illumination-reflectance separation, enhances local contrast and reduces haze

Grayscale to Jet/Hot heatmap pseudocolor mapping, piecewise linear colormap lookup

Reinhard teaching simplified color transfer - RGB channel statistical matching

FFT spectrum analysis and frequency domain filtering - log magnitude spectrum + lowpass/highpass filtering

Motion blur modeling + Wiener deconvolution teaching - line PSF blur to frequency domain Wiener inverse filtering restoration