// The minimum window to contain a given ellipse.
let ellipseWindow (e: Ellipse) =
let cx, cy = roundInt e.Cx, roundInt e.Cy
- let a = int (e.A + 0.5)
+ let a = int (e.A + 0.5f)
cx - a, cy - a, cx + a, cy + a
let w = img.Width
- let w_f = float w
+ let w_f = float32 w
let h = img.Height
- let h_f = float h
+ let h_f = float32 h
// Return 'true' if the point 'p' is owned by e.
// The lines represents all intersections with other ellipses.
tree.Search (searchRegion e)
// We only keep the ellipses touching 'e'.
|> List.choose (fun otherE ->
- match EEOver.EEOverlapArea e otherE with
- | Some (_, px, _) when px.Length > 2 ->
- otherE.Removed <- true
- None
- | Some (area, px, py) when area > 0.0 && px.Length = 2 ->
- Some (otherE, PointD(px.[0], py.[0]), PointD(px.[1], py.[1]))
- | _ ->
+ if e <> otherE
+ then
+ match EEOver.EEOverlapArea e otherE with
+ | Some (_, px, _) when px.Length > 2 ->
+ otherE.Removed <- true
+ None
+ | Some (area, px, py) when area > 0.f && px.Length = 2 ->
+ Some (otherE, PointD(px.[0], py.[0]), PointD(px.[1], py.[1]))
+ | _ ->
+ None
+ else
None )
else
[]
// We reverse the list to get the lower score ellipses first.
let ellipsesWithNeigbors = ellipses |> List.map (fun e -> e, neighbors e) |> List.rev
- // 2) Remove ellipses with a high standard deviation (high contrast).
- // CvInvoke.Normalize(img, img, 0.0, 255.0, CvEnum.NormType.MinMax) // Not needed.
+ // 2) Remove ellipses touching the edges.
+ for e in ellipses do
+ if e.isOutside w_f h_f then e.Removed <- true
- let globalStdDeviation = MathNet.Numerics.Statistics.Statistics.StandardDeviation(seq {
+ // 3) Remove ellipses with a high standard deviation (high contrast).
+ let imgData = img.Data
+ let globalStdDeviation = MathNet.Numerics.Statistics.Statistics.PopulationStandardDeviation(seq {
for y in 0 .. h - 1 do
for x in 0 .. w - 1 do
- yield float img.Data.[y, x, 0] })
+ yield float imgData.[y, x, 0] })
for e in ellipses do
- let minX, minY, maxX, maxY = ellipseWindow e
+ if not e.Removed
+ then
+ let shrinkedE = e.Scale 0.9f
+ let minX, minY, maxX, maxY = ellipseWindow shrinkedE
- let stdDeviation = MathNet.Numerics.Statistics.Statistics.StandardDeviation (seq {
- for y in (if minY < 0 then 0 else minY) .. (if maxY >= h then h - 1 else maxY) do
- for x in (if minX < 0 then 0 else minX) .. (if maxX >= w then w - 1 else maxX) do
- if e.Contains (float x) (float y)
- then
- yield float img.Data.[y, x, 0] })
+ let stdDeviation = MathNet.Numerics.Statistics.Statistics.StandardDeviation (seq {
+ for y in (if minY < 0 then 0 else minY) .. (if maxY >= h then h - 1 else maxY) do
+ for x in (if minX < 0 then 0 else minX) .. (if maxX >= w then w - 1 else maxX) do
+ if shrinkedE.Contains (float32 x) (float32 y)
+ then
+ yield float imgData.[y, x, 0] })
- if stdDeviation > globalStdDeviation * config.Parameters.standardDeviationMaxRatio then
- e.Removed <- true
+ if stdDeviation > globalStdDeviation * config.Parameters.standardDeviationMaxRatio then
+ e.Removed <- true
- // 3) Remove ellipses touching the edges.
- for e in ellipses do
- if e.isOutside w_f h_f then e.Removed <- true
+(*
+ let imgData = img.Data
+ let stdDeviations = [
+ for e in ellipses do
+ if not e.Removed
+ then
+ let shrinkedE = e.Scale 0.9f
+ let minX, minY, maxX, maxY = ellipseWindow shrinkedE
+
+ let stdDeviation = float32 <| MathNet.Numerics.Statistics.Statistics.StandardDeviation (seq {
+ for y in (if minY < 0 then 0 else minY) .. (if maxY >= h then h - 1 else maxY) do
+ for x in (if minX < 0 then 0 else minX) .. (if maxX >= w then w - 1 else maxX) do
+ if shrinkedE.Contains (float32 x) (float32 y)
+ then
+ yield float imgData.[y, x, 0] })
+
+ e.StdDeviation <- stdDeviation
+ yield stdDeviation ]
+
+ // We use Otsu and eliminate some cells only if the curve may be bimodal.
+ // See https://en.wikipedia.org/wiki/Multimodal_distribution#Bimodality_coefficient
+ let skewness, kurtosis = MathNet.Numerics.Statistics.Statistics.PopulationSkewnessKurtosis (stdDeviations |> List.map float)
+ let n = float stdDeviations.Length
+ let bimodalityCoefficient = (skewness ** 2. + 1.) / (kurtosis + 3. * (n - 1.) ** 2. / ((n - 2.) * (n - 3.)))
+
+ if bimodalityCoefficient > 5. / 9.
+ then
+ let hist = ImgTools.histogram stdDeviations 200
+ let thresh, _, _ = ImgTools.otsu hist
+ for e in ellipses do
+ if not e.Removed && e.StdDeviation > thresh
+ then e.Removed <- true
+*)
// 4) Remove ellipses with little area.
let minArea = config.RBCMinArea
let mutable area = 0
for y in (if minY < 0 then 0 else minY) .. (if maxY >= h then h - 1 else maxY) do
for x in (if minX < 0 then 0 else minX) .. (if maxX >= w then w - 1 else maxX) do
- let p = PointD(float x, float y)
+ let p = PointD(float32 x, float32 y)
if pixelOwnedByE p e (neighbors |> List.choose (fun (otherE, p1, p2) -> if otherE.Removed then None else Some (otherE :> Ellipse, Utils.lineFromTwoPoints p1 p2)))
then
area <- area + 1
let elements = new Matrix<byte>(maxY - minY + 1, maxX - minX + 1)
for y in minY .. maxY do
for x in minX .. maxX do
- let p = PointD(float x, float y)
+ let p = PointD(float32 x, float32 y)
if pixelOwnedByE p e (neighbors |> List.choose (fun (otherE, p1, p2) -> if otherE.Removed then None else Some (otherE :> Ellipse, Utils.lineFromTwoPoints p1 p2)))
then
elements.[y-minY, x-minX] <- 1uy
let cellClass =
if float darkStainPixels > config.Parameters.maxDarkStainRatio * (float nbElement) ||
- float stainPixels > config.Parameters.maxStainRatio * (float nbElement) (* ||
- sqrt (((float sumCoords_x) / (float nbElement) - e.Cx) ** 2.0 + ((float sumCoords_y) / (float nbElement) - e.Cy) ** 2.0) > e.A * config.maxOffcenter *)
+ float stainPixels > config.Parameters.maxStainRatio * (float nbElement)
then
Peculiar
elif infectedPixels.Count >= 1