直方图、均值、标准差的正态分布，在数组的某个部分内

Question

I have a lot of different images of a particular object in a folder. I want to set up histograms of these images for Hue, Saturation, and Value. It has worked so far.

On the other hand, I want to fit a normal distribution to the histograms, get the mean and standard deviation. I ran into a problem with the Hue histogram. The histogram shows that it has two parts up to [0-50] and [120-180].

My question is how could I only get the mean and standard deviation in the [120-180] range because, I don't know how to code this.

In the total_hue_list , I tried to count all the occurrences of the different color codes.

import matplotlib.pyplot as plt
import numpy as np
import cv2
import os
from imutils import paths


directory = os.getcwd() + "\Demos/SotetLila"
total_hue_hist = np.zeros((180,))
total_sat_hist = np.zeros((256,))
total_val_hist = np.zeros((256,))


for imagePath in paths.list_images(directory):
    img = cv2.imread(imagePath)
    kernel = np.ones((5, 5), np.uint8)
    blur = cv2.GaussianBlur(img, (5, 5), 0)
    hsv = cv2.cvtColor(blur, cv2.COLOR_BGR2HSV)
    hsv_image = hsv

    hue, sat, val = hsv_image[:, :, 0], hsv_image[:, :, 1], hsv_image[:, :, 2]

    hue_hist, bin_hue = np.histogram(hue, bins=range(181))
    total_hue_hist += hue_hist

    sat_hist, bin_sat = np.histogram(sat, bins=range(257))
    total_sat_hist += sat_hist

    val_hist, bin_val = np.histogram(val, bins=range(257))
    total_val_hist += val_hist



plt.bar(list(range(180)), total_hue_hist)
#plt.bar(list(range(256)), total_sat_hist)
#plt.bar(list(range(256)), total_val_hist)
plt.show()

Answer 1

Just a quick pointer before we start - Hue is cyclic, so assuming your maximal hue is 180, then the hue values 0 and 1 are as close to one another as 180 to 0. I'm not sure from your code if 180 is indeed the max or not, but just putting this fact out there in case it matters.

Back on topic - if I understood correctly, you only want to consider pixels with hue values that are between 120 and 180, and ignore all the other values. If that is indeed the case, then np.histogram has a value designed specifically for this. Quoting the documentation :

range(float, float) , optional The lower and upper range of the bins. If not provided, range is simply (a.min(), a.max()). Values outside the range are ignored. The first element of the range must be less than or equal to the second. range affects the automatic bin computation as well. While bin width is computed to be optimal based on the actual data within range, the bin count will fill the entire range including portions containing no data.

However, if what you really want is just the standard deviation and mean, then you don't really need the histogram, you can just do something similar to this:

def filter_values(values, v_min, v_max):
    # Flatten the array to a single 1D vector
    # See https://numpy.org/doc/stable/reference/generated/numpy.ndarray.flatten.html
    flat = np.asanyarray(values).flatten()

    # Now index the values using "boolean indexing"
    # See https://numpy.org/devdocs/reference/arrays.indexing.html#boolean-array-indexing
    return flat[np.logical_and(flat >= v_min,
                               flat < v_max)]

And you can then compute the mean and standard deviation like so:

filtered_hue = filter_values(hue, v_min=120, v_max=180)
np.mean(filtered_hue), np.std(filtered_hue)

Good luck!