test_opencv.py

# test_opencv.py
# 11-Feb-2018
#
# author: Hernando Vidal, Jr., Tesseract Tech
# for American Museum of Natural History 'Hack the Deep' Hackathon
# 10-11 Feb-2018
##################################################################### 

import sys

import cv2
import matplotlib.pyplot as plt
import numpy as np
import os

from scipy import signal

####################################################################
# dumpJson(sFile, sOut, dX, dY, aOffsets)
#
def dumpJson(sFile,sOut,dX,dY,aOffsets):
    print ("{\n"
        " file:\"%s\",\n"
        " out:\"%s\",\n"
        " width:%d,\n"
        " height:%d,\n"
        % (sFile,sOut,dX, dY))
    sys.stdout.write(" offsets:[")

    if len(aOffsets):
        for i in range(len(aOffsets)-1):
            sys.stdout.write ("%d," % aOffsets[i])

        sys.stdout.write ("%d" % aOffsets[-1])

    print "]\n}"

    return

####################################################################
#
def gammaCorrect(img, fCor):
    img = img / 255.0
    img = cv2.pow(img,fCor)

    return np.uint8(img * 255)

####################################################################
# main()
#   args:   <inputFile> [-n]
#
def main():
    if len(sys.argv) < 2:
        print "Failure"
        exit()
    else:
        sImg = sys.argv[1]

    bLoud = False

    if len(sys.argv) > 2:
        bLoud = True if sys.argv[2] == '-l' else False

    if bLoud:
        print "Image is: ", sImg
  
    # test for output directory
#    if not os.path.isdir("out"):
#        print "Error, output path \"../out\" does not exist"
#        exit()

    imgIn = cv2.imread(sImg)

    if not imgIn.any():
        print "Could not open image"

        exit()

    img = cv2.cvtColor(imgIn,cv2.COLOR_BGR2GRAY)

    dY, dX = img.shape[:2]
    nP = dX * dY

    if bLoud:
        cv2.imshow(sImg,img)
        cv2.waitKey(0)
  
    # establish rough histogram to determine how to scale in brightness/contrast
    hist, bins = np.histogram(img.ravel(),256)

    # sum bins from top to mid in order to determine rough brightness
    nX = 0

    for i in range(255, 127, -1):
        nX += hist[i]

    if bLoud:
        print "Ratio of bright to total:", float(nX) / nP
  
    img = gammaCorrect(img, 2.0)

    if bLoud:
        cv2.imshow("gamma", img)
        cv2.waitKey(0)

    if dX > dY:
        fS = 800.0 / dX
    else:
        fS = 800.0 / dY

    imgCopy = img.copy()

    dXsample = 10
    dX2 = (dX / 2) - dXsample

    accum = np.zeros(dY)
    line = np.zeros(dY)

    if bLoud:
        print "midStart, x:", dX2, " end:", dX2 + dXsample

    for iX in range(dX2-dXsample,dX2+dXsample):
        for iY in range(dY):
            x = img[iY,iX]
            line[iY] = x
            accum[iY] += x

#    plt.plot(line)
#    plt.show()

    cv2.line(imgCopy, (iX, 0), (iX, dY), (255, 255, 255), 1)

    if bLoud:
        cv2.imshow("w line", imgCopy)

    plt.plot(accum)
    plt.title("Raw accum")
    plt.show()

    accum /= (2 * dXsample)

    if bLoud:
        plt.title("Scaled back")
        plt.plot(accum)
        plt.draw()

#    plt.show()
  
    # finds negative peaks in the data
    accum = accum * -1
    peak = signal.find_peaks_cwt(accum,np.arange(1, 25))

    for i in peak:
        if bLoud:
            print "at offset:", i, " value:", accum[i]

    ii = accum[i] * -1

    if ii < 64:
        cv2.line(imgCopy, (0, i), (dX, i), (255, 0, 0), 1)
    else:
        if bLoud:
            print "Point no good"

    if bLoud:
        cv2.imshow('test', imgCopy)
        plt.show()
        cv2.waitKey(0)

    h, t = os.path.split(sImg)

    sOut = t + ".out.jpg"

    cv2.imwrite(sOut, imgCopy)

    if bLoud:
        print peak

    dumpJson(sImg, sOut,dX, dY,peak)

#    print "at x,y", iX, " ", iY, " ", img[iY, iX]

    return

if __name__ == "__main__":
    main()