diff --git a/README.md b/README.md
index 37c676d..be703d2 100644
--- a/README.md
+++ b/README.md
@@ -1,3 +1,18 @@
+## SIFT.py
+- This code will calc the difference between the images and read all the images matrix one by one, the function findSimilarities will apply the sift method saving all images that dont repeat it self and discard the rest.
+
+### Requirements
+- Python 2.7
+- pip install opencv-python==3.4.2.16
+- pip install opencv-contrib-python==3.4.2.16
+- pip install termcolor
+
+### Optionals
+- Change the variable **frame_start** to start the video in a specific frame.
+- Change the variable **IMAGE_DIR** to point your video path.
+- Change the variable **folderTosave** to point to folder that save the frames.
+- Change the variable **thresholdDifference** to change the minimum threshold at compare two frames.
+
 ## differenceBetweenImages.py
 - Calc the difference between two frames, and save the next frame that difference is less than a threshold
 
diff --git a/SIFT.py b/SIFT.py
new file mode 100644
index 0000000..f74f2ab
--- /dev/null
+++ b/SIFT.py
@@ -0,0 +1,119 @@
+#!/usr/bin/env python
+# coding: utf-8
+
+# In[1]:
+
+
+#pip install opencv-python==3.4.2.16
+#pip install opencv-contrib-python==3.4.2.16
+#pip install termcolor
+
+
+# In[1]:
+
+
+from PIL import Image
+import glob
+import cv2
+from termcolor import colored
+from os.path import join
+
+
+# In[3]:
+
+
+IMAGE_DIR = join('zip10')
+images = glob.glob(f'{IMAGE_DIR}/*.jpg')
+folderTosave = "Tosave"
+print(f'Found {len(images)} images')
+thresholdDifference = 5
+frame_start = -1
+
+
+# In[5]:
+
+
+def main():
+    print("Image Path: %s" % IMAGE_DIR)
+    for image in images:
+        img = Image.open(f'{image}')
+        width, height = img.size[:2]
+        width, height = int(width), int(height)
+    
+    count = -1
+    origin = 0
+    frameOriginal = 0
+    ############################
+    #Frame Size
+    yframeFrom = 0
+    yframeTo = height
+
+    xFrameFrom = 0
+    xFrameTo =  width
+
+    for img in glob.glob(f'{IMAGE_DIR}/*.jpg'):
+        count+=1
+        cap = cv2.imread(img)
+        ret = cap
+        print("Frame: "+str(count))
+        if count == frame_start+1:
+            frameOriginal = cap[yframeFrom:yframeTo, xFrameFrom:xFrameTo]
+            pathToSave = join(folderTosave, "%s.jpg" % (count))
+            cv2.imwrite(pathToSave, frameOriginal)
+        else:
+            if not ret.any():
+                break
+            frameCompare = cap[yframeFrom:yframeTo,xFrameFrom:xFrameTo]
+            #cv2.imshow("Original", frameOriginal)
+            #cv2.imshow("Comparacao",frameCompare)
+            #cv2.waitKey(0)
+            #cv2.destroyAllWindows()
+            try:
+                difference = findSimilarities(frameOriginal,frameCompare)
+                if(difference > thresholdDifference):
+                    print(colored("Accepted",'green'))
+                    pathToSave = join(folderTosave, "%s.jpg" % (count))
+                    cv2.imwrite(pathToSave, frameCompare)
+                    frameOriginal = frameCompare.copy()
+                else:
+                    print("Diff expected, less than %s" % str(thresholdDifference))
+                    print(colored("Denied",'red'))
+                if cv2.waitKey(1) & 0xFF == ord('q'):
+                    break
+                print("\n")
+            except:
+                continue
+        
+def findSimilarities(original, compare): 
+    sift = cv2.xfeatures2d.SIFT_create()
+    kp_1, desc_1 = sift.detectAndCompute(original, None)
+    kp_2, desc_2 = sift.detectAndCompute(compare, None)
+
+    index_params = dict(algorithm=0, trees=5)
+    search_params = dict()
+    flann = cv2.FlannBasedMatcher(index_params, search_params)
+
+    matches = flann.knnMatch(desc_1, desc_2, k=2)
+
+    good_points = []
+    ratio = 0.6
+    for m, n in matches:
+        if m.distance < ratio*n.distance:
+            good_points.append(m)
+    lGoodPoints = len(good_points)
+
+    #TO SEE THE FRAME COMPARED UNCOMMENT THE LINES BELOW
+    #result = cv2.drawMatches(original, kp_1, compare, kp_2, good_points, None)
+    #cv2.imshow("result", result)
+    #cv2.waitKey(0)
+    #cv2.destroyAllWindows()
+    return lGoodPoints
+if __name__ == '__main__':
+    main()
+
+
+# In[ ]:
+
+
+
+