Jsrtaa is a Java library for image processing, focused on being a foundational library for scaling, denoising, and super-resolution.

Why would you want to use Jsrtaa?

• Support development of Java-based image processing algorithms
• Free for non-commericial use, unlike many other SR applications
• Support for colour images and animated GIFs
• Has a command-line interface for scripting
• Implements some algorithms only available in the scientific literature or in other languages

Overview of current capabilities of Jsrtaa:

• Linear convolutional filters
  • Gaussian & box blur
  • Discrete sharpening
  • Bicubic kernel
• Nonlinear filters
  • Median
  • Tristate median
• Interpolation
  • Nearest neighbour
  • Bilinear
  • Bicubic polynomial
  • Lanczos kernel
  • Bicubic Akima spline
• Denoising
  • Non-local Means
  • Whole-image singular value removal
  • Bilateral neighbourhood filtering
• Anisotropic diffusion
  • Perona-Malik non-linear anisotropic diffusion
  • Curvature-driven non-linear anisotropic diffusion
• Noise
  • Addition: Gaussian, uniform, or poisson additive noise & impulse noise
  • Level Estimation: Canny estimator, Gaussian noise variance, & non-parametric noise variance
• Transformation
  • Logarithmic normalization
  • Finite difference gradient, Sobel gradient, & Laplacian image
  • Fourier space image
  • Mean & Gaussian curvature
  • Greyscale: pixel equalization or luminance-preserving CIE1931 colorimetric conversion
• Metrics
  • Comparison: RMSE, PSNR, Jensen-Shannon divergence
  • Statistics: median absolute deviation of the gradient, information-theoretic entropy
• Super-resolution
  • Locally adaptive zooming (LAZA)