我需要一种算法,可以确定两个图像是否“相似”,并识别颜色、亮度、形状等方面的相似模式。 我可能需要一些指针,了解人脑用于“分类”图像的参数。
我已经看过基于哈斯多夫匹配的方法,但似乎主要用于匹配变换后的对象和形状的模式。
我需要一种算法,可以确定两个图像是否“相似”,并识别颜色、亮度、形状等方面的相似模式。 我可能需要一些指针,了解人脑用于“分类”图像的参数。
我已经看过基于哈斯多夫匹配的方法,但似乎主要用于匹配变换后的对象和形状的模式。
我曾采用过一种类似的方法,通过使用小波变换将图像分解为特征。
我的方法是从每个转换后的信道中选择最重要的n个系数,并记录它们的位置。这是通过根据abs(power)对(power,location)元组列表进行排序来完成的。相似的图像将在相同的位置具有显著的系数。
我发现最好将图像转换为YUV格式,这有效地允许您在形状(Y通道)和颜色(UV通道)上加权相似性。
您可以在mactorii中找到上述实现,不幸的是,我没有像应该做的那样多地工作在这上面 :-)。
另一种方法是,我的一些朋友用了一个令人惊讶的好结果,就是将图像简单地调整大小到4x4像素,然后将其存储为您的特征。可以通过计算两个图像之间的曼哈顿距离来评估它们的相似程度,使用相应像素。我没有它们执行缩放的详细信息,所以您可能需要尝试使用各种可用于此任务的算法来找到适合您的算法。
pHash可能会引起您的兴趣。
感知哈希(perceptual hash) n. 音频、视频或图像文件的指纹,基于其包含的音频或视觉内容的数学计算。与依赖于输入微小变化引起输出剧烈变化的密码哈希函数不同,如果输入在视觉上或听觉上相似,则感知哈希是“接近”的。
我曾经使用SIFT在不同的图像中重新检测相同的对象。它非常强大但也相当复杂,可能有些过度。如果这些图像应该是相似的,则基于两个图像之间差异的一些简单参数可以告诉你很多信息。以下是一些提示:
from __future__ import absolute_import, division, print_function
"""
This is a modification of the classify_images.py
script in Tensorflow. The original script produces
string labels for input images (e.g. you input a picture
of a cat and the script returns the string "cat"); this
modification reads in a directory of images and
generates a vector representation of the image using
the penultimate layer of neural network weights.
Usage: python classify_images.py "../image_dir/*.jpg"
"""
# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Simple image classification with Inception.
Run image classification with Inception trained on ImageNet 2012 Challenge data
set.
This program creates a graph from a saved GraphDef protocol buffer,
and runs inference on an input JPEG image. It outputs human readable
strings of the top 5 predictions along with their probabilities.
Change the --image_file argument to any jpg image to compute a
classification of that image.
Please see the tutorial and website for a detailed description of how
to use this script to perform image recognition.
https://tensorflow.org/tutorials/image_recognition/
"""
import os.path
import re
import sys
import tarfile
import glob
import json
import psutil
from collections import defaultdict
import numpy as np
from six.moves import urllib
import tensorflow as tf
FLAGS = tf.app.flags.FLAGS
# classify_image_graph_def.pb:
# Binary representation of the GraphDef protocol buffer.
# imagenet_synset_to_human_label_map.txt:
# Map from synset ID to a human readable string.
# imagenet_2012_challenge_label_map_proto.pbtxt:
# Text representation of a protocol buffer mapping a label to synset ID.
tf.app.flags.DEFINE_string(
'model_dir', '/tmp/imagenet',
"""Path to classify_image_graph_def.pb, """
"""imagenet_synset_to_human_label_map.txt, and """
"""imagenet_2012_challenge_label_map_proto.pbtxt.""")
tf.app.flags.DEFINE_string('image_file', '',
"""Absolute path to image file.""")
tf.app.flags.DEFINE_integer('num_top_predictions', 5,
"""Display this many predictions.""")
# pylint: disable=line-too-long
DATA_URL = 'http://download.tensorflow.org/models/image/imagenet/inception-2015-12-05.tgz'
# pylint: enable=line-too-long
class NodeLookup(object):
"""Converts integer node ID's to human readable labels."""
def __init__(self,
label_lookup_path=None,
uid_lookup_path=None):
if not label_lookup_path:
label_lookup_path = os.path.join(
FLAGS.model_dir, 'imagenet_2012_challenge_label_map_proto.pbtxt')
if not uid_lookup_path:
uid_lookup_path = os.path.join(
FLAGS.model_dir, 'imagenet_synset_to_human_label_map.txt')
self.node_lookup = self.load(label_lookup_path, uid_lookup_path)
def load(self, label_lookup_path, uid_lookup_path):
"""Loads a human readable English name for each softmax node.
Args:
label_lookup_path: string UID to integer node ID.
uid_lookup_path: string UID to human-readable string.
Returns:
dict from integer node ID to human-readable string.
"""
if not tf.gfile.Exists(uid_lookup_path):
tf.logging.fatal('File does not exist %s', uid_lookup_path)
if not tf.gfile.Exists(label_lookup_path):
tf.logging.fatal('File does not exist %s', label_lookup_path)
# Loads mapping from string UID to human-readable string
proto_as_ascii_lines = tf.gfile.GFile(uid_lookup_path).readlines()
uid_to_human = {}
p = re.compile(r'[n\d]*[ \S,]*')
for line in proto_as_ascii_lines:
parsed_items = p.findall(line)
uid = parsed_items[0]
human_string = parsed_items[2]
uid_to_human[uid] = human_string
# Loads mapping from string UID to integer node ID.
node_id_to_uid = {}
proto_as_ascii = tf.gfile.GFile(label_lookup_path).readlines()
for line in proto_as_ascii:
if line.startswith(' target_class:'):
target_class = int(line.split(': ')[1])
if line.startswith(' target_class_string:'):
target_class_string = line.split(': ')[1]
node_id_to_uid[target_class] = target_class_string[1:-2]
# Loads the final mapping of integer node ID to human-readable string
node_id_to_name = {}
for key, val in node_id_to_uid.items():
if val not in uid_to_human:
tf.logging.fatal('Failed to locate: %s', val)
name = uid_to_human[val]
node_id_to_name[key] = name
return node_id_to_name
def id_to_string(self, node_id):
if node_id not in self.node_lookup:
return ''
return self.node_lookup[node_id]
def create_graph():
"""Creates a graph from saved GraphDef file and returns a saver."""
# Creates graph from saved graph_def.pb.
with tf.gfile.FastGFile(os.path.join(
FLAGS.model_dir, 'classify_image_graph_def.pb'), 'rb') as f:
graph_def = tf.GraphDef()
graph_def.ParseFromString(f.read())
_ = tf.import_graph_def(graph_def, name='')
def run_inference_on_images(image_list, output_dir):
"""Runs inference on an image list.
Args:
image_list: a list of images.
output_dir: the directory in which image vectors will be saved
Returns:
image_to_labels: a dictionary with image file keys and predicted
text label values
"""
image_to_labels = defaultdict(list)
create_graph()
with tf.Session() as sess:
# Some useful tensors:
# 'softmax:0': A tensor containing the normalized prediction across
# 1000 labels.
# 'pool_3:0': A tensor containing the next-to-last layer containing 2048
# float description of the image.
# 'DecodeJpeg/contents:0': A tensor containing a string providing JPEG
# encoding of the image.
# Runs the softmax tensor by feeding the image_data as input to the graph.
softmax_tensor = sess.graph.get_tensor_by_name('softmax:0')
for image_index, image in enumerate(image_list):
try:
print("parsing", image_index, image, "\n")
if not tf.gfile.Exists(image):
tf.logging.fatal('File does not exist %s', image)
with tf.gfile.FastGFile(image, 'rb') as f:
image_data = f.read()
predictions = sess.run(softmax_tensor,
{'DecodeJpeg/contents:0': image_data})
predictions = np.squeeze(predictions)
###
# Get penultimate layer weights
###
feature_tensor = sess.graph.get_tensor_by_name('pool_3:0')
feature_set = sess.run(feature_tensor,
{'DecodeJpeg/contents:0': image_data})
feature_vector = np.squeeze(feature_set)
outfile_name = os.path.basename(image) + ".npz"
out_path = os.path.join(output_dir, outfile_name)
np.savetxt(out_path, feature_vector, delimiter=',')
# Creates node ID --> English string lookup.
node_lookup = NodeLookup()
top_k = predictions.argsort()[-FLAGS.num_top_predictions:][::-1]
for node_id in top_k:
human_string = node_lookup.id_to_string(node_id)
score = predictions[node_id]
print("results for", image)
print('%s (score = %.5f)' % (human_string, score))
print("\n")
image_to_labels[image].append(
{
"labels": human_string,
"score": str(score)
}
)
# close the open file handlers
proc = psutil.Process()
open_files = proc.open_files()
for open_file in open_files:
file_handler = getattr(open_file, "fd")
os.close(file_handler)
except:
print('could not process image index',image_index,'image', image)
return image_to_labels
def maybe_download_and_extract():
"""Download and extract model tar file."""
dest_directory = FLAGS.model_dir
if not os.path.exists(dest_directory):
os.makedirs(dest_directory)
filename = DATA_URL.split('/')[-1]
filepath = os.path.join(dest_directory, filename)
if not os.path.exists(filepath):
def _progress(count, block_size, total_size):
sys.stdout.write('\r>> Downloading %s %.1f%%' % (
filename, float(count * block_size) / float(total_size) * 100.0))
sys.stdout.flush()
filepath, _ = urllib.request.urlretrieve(DATA_URL, filepath, _progress)
print()
statinfo = os.stat(filepath)
print('Succesfully downloaded', filename, statinfo.st_size, 'bytes.')
tarfile.open(filepath, 'r:gz').extractall(dest_directory)
def main(_):
maybe_download_and_extract()
if len(sys.argv) < 2:
print("please provide a glob path to one or more images, e.g.")
print("python classify_image_modified.py '../cats/*.jpg'")
sys.exit()
else:
output_dir = "image_vectors"
if not os.path.exists(output_dir):
os.makedirs(output_dir)
images = glob.glob(sys.argv[1])
image_to_labels = run_inference_on_images(images, output_dir)
with open("image_to_labels.json", "w") as img_to_labels_out:
json.dump(image_to_labels, img_to_labels_out)
print("all done")
if __name__ == '__main__':
tf.app.run()
http://www.hackerfactor.com/blog/index.php?/archives/432-Looks-Like-It.html
有相关研究使用Kohonen神经网络/自组织映射。
无论是更加学术的系统(例如Google的PicSOM),还是较少学术性的演示文稿(例如http://www.generation5.org/content/2004/aiSomPic.asp,可能不适用于所有工作环境),都存在。
计算缩小版本(例如:6x6像素)像素颜色值差的平方和非常有效。相同的图像产生0,相似的图像产生小数字,不同的图像产生大数字。
其他人提出的先分解成YUV的想法听起来很有趣 - 虽然我的想法很好用,但我希望我的图像被计算为“不同”,以便得到正确的结果 - 即使从色盲观察者的角度也是如此。