Computer Vision Lab Algorithms
Algorithms
56 results
- Converting to greyscaleGrayscale conversion (ITU-R BT.601)Converting an RGB colour image into a single intensity channel with a weighted sum of the three channels.Image processingIn the browser
- Adjusting brightness and contrastA per-pixel linear transform that multiplies the intensity by a contrast factor and adds a brightness offset.Image processingIn the browser
- Gamma correctionA non-linear transform that raises the normalised intensity to the power γ.Image processingIn the browser
- Linear (min-max) normalisationMin-max normalizationRescaling the values to a fixed range such as 0–255 or 0–1.Image processingIn the browser
- Image resizing and interpolationChanging the image dimensions by estimating the new pixel values from their neighbours.Image processingIn the browser
- Gaussian smoothingGaussian blurConvolving the image with a Gaussian kernel to reduce high-frequency noise.Image processingIn the browser
- The mean filterReplacing each pixel with the mean of its neighbours within a square window.Image processingIn the browser
- The median filterReplacing each pixel with the statistical median of the values in its window.Image processingIn the browser
- The bilateral filterEdge-preserving smoothing that weights neighbours by both spatial distance and intensity difference.Image processingIn the browser
- Sharpening with unsharp maskingAdding the difference between the image and its smoothed copy to strengthen the edges.Image processingIn the browser
- The Sobel operatorApproximating the first derivative of the image with two 3×3 kernels, horizontal and vertical.Image processingIn the browser
- The Laplacian operatorApproximating the isotropic second derivative.Image processingIn the browser
- The Canny edge detectorA multi-stage algorithm that gives thin, connected edges.Image processingIn the browser
- Fixed thresholdGlobal thresholdingConverting the image to binary by comparing each pixel with a single threshold.Image processingIn the browser
- Otsu's thresholding methodChoosing the threshold automatically by maximising the variance between the two classes in the histogram.Image processingIn the browser
- Adaptive thresholdAdaptive thresholdingComputing a different threshold for each pixel from the mean of its neighbourhood minus a constant.Image processingIn the browser
- Morphological operationsMathematical morphologyOperations on shapes based on a structuring element: erosion, dilation, opening and closing.Image processingIn the browser
- Histogram equalisationHistogram equalizationRedistributing the intensities to flatten the histogram.Image processingIn the browser
- Contrast-limited adaptive histogram equalisationCLAHELocal equalisation on small tiles, with histogram clipping to avoid amplifying noise.Medical imagesIn the browser
- 2D convolutionSliding a small kernel over the image and computing the sum of products at each position.Convolution and neural networksIn the browser
- PoolingShrinking the feature map by taking the maximum or the mean in each window.Convolution and neural networksIn the browser
- The ReLU activation functionRectified Linear UnitAn activation function that zeroes negative values and passes positive ones.Convolution and neural networksIn the browser
- Batch normalisationBatch normalizationNormalising a layer's outputs with the batch mean and variance, then rescaling them with learned coefficients.Convolution and neural networksConcept and explanation
- The Softmax functionConverting scores (logits) into a probability distribution.Image classificationIn the browser
