Automated inverse telecine conversion

A method of performing an inverse telecine process on a sequence of non-interlaced video frames that has been converted from a file format to a video format in accordance with a 3:2 telecine process, the method comprising:

• receiving a sequence of video farms at a video frame rate, where the sequence includes duplicate frames that are the result of the 3:2 telecine process;
• determining a degree of difference between each frame and its adjacent frame within at least a portion of the video sequence, and storing those difference values;
• making a first selection of difference values, wherein the difference values selected are computed 5 frame intervals apart;
• making a second selection of difference values, where at least a portion of the difference values associated with difference values other than those in the it selection;
• analyzing the difference values associated with the first selection and the difference values associated with the second selection, and locating at least a first duplicate frame;
• deleting the first duplicate frame;
• tracking a duplicate fame position of a previously detected telecine pattern;
• receiving a portion of the sequence of frames wherein the frames vary by less than a predetermined amount;
• and
• attempting to detect and upon a failure to detect a telecine pattern in a collection, deleting a frame when the frame corresponds to a frame position consistent with the previously detected telecine pattern.

A method of processing a sequence of interlaced video frames that has been converted from a film format to a video format in accordance with a telecine process, where the method transforms the sequence back to the film format without user intervention, the method comprising:

• receiving a sequence of video frames at a video frame rate, where a video frame from the sequence includes both an even field and an odd field, where the sequence includes duplicate even and odd fields generated by the telecine process;
• determining a degree of difference between even and odd fields of adjacent frames within at least a portion of the sequence, and storing these difference values;
• making a first selection of the difference values between even fields of adjacent frames, where the difference values selected are computed 5 frame intervals apart;
• making a second selection of at least a portion of the difference values between even fields associated with even field difference values other than those in the first selection;
• making a third selection of the difference values between odd fields of adjacent fields, where the difference values selected are computed 5 frame intervals apart, where the 5 frame intervals for the third selection are offset by 2 frame intervals from the 5 frame intervals for the first selection;
• making a fourth selection of at least a portion of the difference values between odd fields associated with the odd field difference values other than those in the third selection;
• analyzing the difference values associated with the first selection, second selection, third selection, and the fourth selection, and locating at least a first duplicate frame;
• and
• deleting the first duplicate frame.

A method of maintaining an inverse telecine process when a difference between frames of at least a portion a telecine processed sequence of frames is less that a predetermined amount, the method comprising:

• detecting a telecine pattern in a first portion of the telecine processed sequence of frames;
• attempting to detect a telecine pattern in a second portion of the telecine processed sequence of frames;
• upon failing to detect a telecine pattern:
• detecting that the adjacent frames in the second portion are similar to each other;
• predicting a position for a duplicate frame that is the result of a telecine process, where the prediction is based on the telecine pattern observed in the first portion and a passage of time;
• and
• deleting a frame from the second portion when the frame corresponds to the predicted position.