Issues

• imgproc: hough_circles differs from OpenCV 4.10.0: not the HOUGH_GRADIENT algorithm, max_radius 0 finds nothing, panics on dp = 0

  bugimgproc-module
  webarkit/purecv#248 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: hough_lines_p differs from OpenCV 4.10.0: not deterministic (OpenCV seeds cv::RNG with -1)

  bugimgproc-module
  webarkit/purecv#247 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: hough_lines differs from OpenCV 4.10.0: order of equal-vote lines, numrho rounding

  bugimgproc-module
  webarkit/purecv#246 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: corner_sub_pix differs from OpenCV 4.10.0: no Gaussian window, Sobel instead of getRectSubPix differences, no reset of diverging corners

  bugimgproc-module
  webarkit/purecv#245 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: good_features_to_track differs from OpenCV 4.10.0: order of equal responses, no mask/gradient_size

  bugimgproc-module
  webarkit/purecv#244 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: corner_harris/corner_min_eigen_val/corner_eigen_vals_and_vecs/pre_corner_detect differ from OpenCV 4.10.0: responses not normalised, pre_corner_detect uses derivatives of derivatives

  bugimgproc-module
  webarkit/purecv#243 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: canny differs from OpenCV 4.10.0: two-pixel-wide edges, border pixels never marked, threshold ties, aperture 7

  bugimgproc-module
  webarkit/purecv#242 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: Clahe differs from OpenCV 4.10.0: padding when only one dimension is not divisible by the tile grid

  bugimgproc-module
  webarkit/purecv#241 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: calc_back_project differs from OpenCV 4.10.0: output always f32 clamped to [0, 255] instead of the images' depth

  bugimgproc-module
  webarkit/purecv#240 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: HistCompMethods differs from OpenCV 4.10.0: Intersection, Bhattacharyya and ChiSqrAlt discriminants

  bugimgproc-module
  webarkit/purecv#239 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: pyr_down/pyr_up differ from OpenCV 4.10.0: float outputs rounded to integers, pyr_up edge rule and odd sizes, invalid sizes and borders accepted

  bugimgproc-module
  webarkit/purecv#238 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: remap/warp_perspective differ from OpenCV 4.10.0: coordinates not quantised to 1/32 px, no fixed-point u8 blend, results depend on --features simd, BORDER_TRANSPARENT, no WARP_INVERSE_MAP

  bugimgproc-moduleSIMD
  webarkit/purecv#237 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: resize differs from OpenCV 4.10.0: float outputs rounded to integers, u8 not OpenCV's fixed-point INTER_LINEAR, no fx/fy/interpolation

  bugimgproc-module
  webarkit/purecv#236 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: threshold differs from OpenCV 4.10.0: integer thresh not floored (and returned unfloored), maxval not rounded, simd path clamps thresh, OTSU/TRIANGLE missing

  bugimgproc-moduleSIMD
  webarkit/purecv#235 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: erode/dilate/morphology_ex panic on multi-channel images with --features simd

  bugimgproc-moduleSIMD
  webarkit/purecv#234 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: get_structuring_element differs from OpenCV 4.10.0: ellipse rows truncate instead of rounding (and erode/dilate reject an empty kernel)

  bugimgproc-module
  webarkit/purecv#233 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: get_sobel_kernels/sobel/scharr/laplacian differ from OpenCV 4.10.0: ksize 1 and 7-31 unsupported, invalid dx/dy accepted, no normalize

  bugimgproc-module
  webarkit/purecv#232 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: bilateral_filter differs from OpenCV 4.10.0: square window, L2 colour distance, no sigma/radius clamps, any channel count

  bugimgproc-module
  webarkit/purecv#231 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: median_blur differs from OpenCV 4.10.0: reflect-101 border instead of replicate, ksize 1 rejected, unsupported depth/size combinations accepted

  bugimgproc-module
  webarkit/purecv#230 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• imgproc: get_gaussian_kernel/gaussian_blur differ from OpenCV 4.10.0: sigma <= 0 tables, even sizes, ksize from sigma, panic on non-square ksize with sigma2 = 0, no 8U bit-exact path

  bugimgproc-module
  webarkit/purecv#229 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• core: CV_LOG2 differs from OpenCV 4.10.0: log2(e) instead of ln 2

  bugcore-module
  webarkit/purecv#226 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• core: use cv::RNG as the default generator for randu/randn/rand_shuffle (keep xoshiro256** as an option; randn must saturate integer samples)

  core-moduleenhancement
  webarkit/purecv#225 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• core: idft differs from OpenCV 4.10.0: DFT_SCALE is always applied

  bugcore-module
  webarkit/purecv#224 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• core: dft differs from OpenCV 4.10.0: real input gives complex output where OpenCV packs CCS (keep complex behind DFT_COMPLEX_OUTPUT)

  bugcore-module
  webarkit/purecv#223 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0

• core: copy_make_border differs from OpenCV 4.10.0: border codes do not follow BorderTypes, oversized borders panic

  bugcore-module
  webarkit/purecv#222 · Filter by author kalwalt opened on Oct 19, 2026 · Parity audit: purecv ↔ OpenCV 4.10.0