op MSER

OpenCV MSER wrapper with FireSight filtering extensions.

Pipeline

{"op":"MSER", "color":[255, 0, 255], "minArea":1100, "maxArea":1200}

• delta delta, in the code, it compares (size_{i}-size_{i-delta})/size_{i-delta}. Larger delta yields fewer images. Default 5
• minArea prune the blobs whose area is smaller than minArea. Default 60
• maxArea prune the blobs whose area is bigger than maxArea. Default 14400
• maxVariation prune the area have simliar size to its children. Default 0.25
• minDiversity trace back to cut off mser with diversity < min_diversity. Default 0.2
• maxEvolution for color image, the evolution steps. Default 200
• areaThreshold the area threshold to cause re-initialize. Default 1.01
• minMargin ignore too small margin. Default 0.003
• edgeBlurSize the aperture size for edge blur. Default 5
• mask If provided, a JSON object (x,y,width,height) specifies MSER mask rectangle. By default, entire image is processed.

The following pipeline parameters are exclusive to FireSight:

• color if provided, matched regions will be colored with BGR Scalar having 1-4 values. For alternating colors, use: [-1,-1,-1,-1]
• detect keypoints adds a KeyPoint JSON object to the stage model for each detected MSER region. Each KeyPoint is determined from eigenvectors and covariance matrix of region. The size field of the KeyPoint is calculated from the number of points in the region: size = 2*sqrt(region.size()/π)
• detect rects adds a RotatedRect JSON object to the stage model for each detected MSER region. Each RotatedRect is calculated by minAreaRect()

Stage Model

If "detect":"keypoints" is specified, a stage model KeyPoint is added for each MSER region detected. The following model corresponds with the MSER with keypoints example below:

 "mser":{
    "keypoints":[
      {
        "pt.x":209.50672912597656,
        "pt.y":71.5,
        "size":38.892269134521484,
        "angle":359.97735595703125
      },
      {
        "pt.x":121.55968475341797,
        "pt.y":72.026527404785156,
        "size":37.947780609130859,
        "angle":359.83328247070312
      }
    ]
  }

If "detect":"rects" is specified, the stage model is updated with a JSON object for each region's minAreaRect() RotatedRect:

{
    "rects":[
      {
        "x":209.0,
        "y":71.0,
        "width":54.0,
        "height":22.0,
        "angle":0.0
      },
      {
        "x":121.5,
        "y":72.0,
        "width":53.0,
        "height":20.0,
        "angle":-0.0
      }
    ]
  }

Examples

In the following examples, output image is written to target/MSER.jpg

MSER with "detect":"keypoints" pipeline

firesight -i img/ass_place_phj.jpg -p json/MSER_phj.json -o target/MSER.jpg

MSER with Mask pipeline

firesight -i img/ass_place_phf.jpg -p json/MSER_phf.json -o target/MSER.jpg

Grayscale MSER with Mask pipeline

firesight -i img/ass_place_phh.jpg -p json/MSER_phh.json -o target/MSER.jpg

MSER with "detect":"keypoints" and "detect":"rects" pipeline

firesight -i img/ass_place_phj.jpg -p json/MSER_phj2.json -o target/MSER_phj2.jpg